tracing: Allow system call tracepoints to handle page faults
Mathieu Desnoyers
reviewerv4 cover tracing: Allow system call tracepoints to handle page faults
Wire up the system call tracepoints with Tasks Trace RCU to allow the ftrace, perf, and eBPF tracers to handle page faults.
This series does the initial wire-up allowing tracers to handle page faults, but leaves out the actual handling of said page faults as future work.
This series was compile and runtime tested with ftrace and perf syscall tracing and raw syscall tracing, adding a WARN_ON_ONCE() in the generated code to validate that the intended probes are used for raw syscall tracing. The might_fault() added within those probes validate that they are called from a context where handling a page fault is OK.
The eBPF people have been needing this feature for quite a while:
https://lore.kernel.org/lkml/[email protected]/
This series replaces the "Faultable Tracepoints v6" series found at [1].
This has been rebased on trace/core commit:
1537519062e2 (trace/core) tracepoint: Remove SRCU protection
The changes since v3 take care of feedback from Steven.
Thanks,
Mathieu
Mathieu Desnoyers (8):
tracing: Declare system call tracepoints with TRACE_EVENT_SYSCALL tracing/ftrace: disable preemption in syscall probe tracing/perf: disable preemption in syscall probe tracing/bpf: disable preemption in syscall probe
tracing: Allow system call tracepoints to handle page faults tracing/ftrace: Add might_fault check to syscall probes tracing/perf: Add might_fault check to syscall probes tracing/bpf: Add might_fault check to syscall probes
include/linux/tracepoint.h | 71 ++++++++++++++++++++++++++------- include/trace/bpf_probe.h | 14 +++++++ include/trace/define_trace.h | 5 +++ include/trace/events/syscalls.h | 4 +- include/trace/perf.h | 44 ++++++++++++++++++-- include/trace/trace_events.h | 62 ++++++++++++++++++++++++++-- init/Kconfig | 1 + kernel/trace/trace_syscalls.c | 28 +++++++++++++ 8 files changed, 205 insertions(+), 24 deletions(-)
Signature · 1 line
2.39.2
Mathieu Desnoyers
reviewerPATCH v4 1/8 tracing: Declare system call tracepoints with TRACE_EVENT_SYSCALL
In preparation for allowing system call tracepoints to handle page faults, introduce TRACE_EVENT_SYSCALL to declare the sys_enter/sys_exit tracepoints.
Move the common code between __DECLARE_TRACE and __DECLARE_TRACE_SYSCALL into __DECLARE_TRACE_COMMON.
This change is not meant to alter the generated code, and only prepares the following modifications.
---
Changes since v0:
- Fix allnoconfig build by adding __DECLARE_TRACE_SYSCALL define in CONFIG_TRACEPOINTS=n case.
- Rename unregister_trace_sys_{enter,exit} to unregister_trace_syscall_sys_{enter,exit} for symmetry with register.
- Add emit trace_syscall_##name##_enabled for syscall tracepoints rather than trace_##name##_enabled, so it is in sync with the rest of the naming.
Changes since v1:
- Rebase on top of Steven's tracepoint patches removing rcuidle variant and srcu protection.
Changes since v2:
- Remove introduction of trace_syscall_ prefix to minimize code churn. Changes since v3:
- Rebase on top of trace/core, replace s/static_key_false/static_branch_unlikely/ and adapt tabs.
--- include/linux/tracepoint.h | 53 +++++++++++++++++++++++++-------- include/trace/bpf_probe.h | 3 ++ include/trace/define_trace.h | 5 ++++ include/trace/events/syscalls.h | 4 +-- include/trace/perf.h | 3 ++ include/trace/trace_events.h | 28 +++++++++++++++++ 6 files changed, 81 insertions(+), 15 deletions(-)
include/linux/tracepoint.h +40 -13
@@ -197,7 +197,7 @@ static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p)* it_func[0] is never NULL because there is at least one element in the array * when the array itself is non NULL. */ -#define __DO_TRACE (name, args, cond) \ +#define __DO_TRACE (name, args, cond, syscall) \ do { \ int __maybe_unused __idx = 0; \ \
@@ -222,21 +222,10 @@ static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p)* site if it is not watching, as it will need to be active when the * tracepoint is enabled. */ -#define __DECLARE_TRACE (name, proto, args, cond, data_proto) \ +#define __DECLARE_TRACE_COMMON (name, proto, args, cond, data_proto) \ extern int __traceiter_##name(data_proto); \ DECLARE_STATIC_CALL(tp_func_##name, __traceiter_##name); \ extern struct tracepoint __tracepoint_##name; \ - static inline void trace_##name(proto) \ - { \ - if (static_branch_unlikely(&__tracepoint_##name.key)) \ - __DO_TRACE(name, \ - TP_ARGS(args), \ - TP_CONDITION(cond)); \ - if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) { \ - WARN_ONCE(!rcu_is_watching(), \ - "RCU not watching for tracepoint"); \ - } \ - } \ static inline int \ register_trace_##name(void (*probe)(data_proto), void *data) \ { \
@@ -266,6 +255,34 @@ static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p) return static_branch_unlikely(&__tracepoint_##name.key);\ } +#define __DECLARE_TRACE (name, proto, args, cond, data_proto) \ + __DECLARE_TRACE_COMMON(name, PARAMS(proto), PARAMS(args), cond, PARAMS(data_proto)) \ + static inline void trace_##name(proto) \ + { \ + if (static_branch_unlikely(&__tracepoint_##name.key)) \ + __DO_TRACE(name, \ + TP_ARGS(args), \ + TP_CONDITION(cond), 0); \ + if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) { \ + WARN_ONCE(!rcu_is_watching(), \ + "RCU not watching for tracepoint"); \ + } \ + } + +#define __DECLARE_TRACE_SYSCALL (name, proto, args, cond, data_proto) \ + __DECLARE_TRACE_COMMON(name, PARAMS(proto), PARAMS(args), cond, PARAMS(data_proto)) \ + static inline void trace_##name(proto) \ + { \ + if (static_branch_unlikely(&__tracepoint_##name.key)) \ + __DO_TRACE(name, \ + TP_ARGS(args), \ + TP_CONDITION(cond), 1); \ + if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) { \ + WARN_ONCE(!rcu_is_watching(), \ + "RCU not watching for tracepoint"); \ + } \ + } +/* * We have no guarantee that gcc and the linker won't up-align the tracepoint * structures, so we create an array of pointers that will be used for iteration
@@ -348,6 +365,8 @@ static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p)return false; \ } +#define __DECLARE_TRACE_SYSCALL __DECLARE_TRACE +#define DEFINE_TRACE_FN (name, reg, unreg, proto, args)#define DEFINE_TRACE (name, proto, args)#define EXPORT_TRACEPOINT_SYMBOL_GPL (name)
@@ -409,6 +428,11 @@ static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p) cpu_online(raw_smp_processor_id()) && (PARAMS(cond)), \ PARAMS(void *__data, proto)) +#define DECLARE_TRACE_SYSCALL (name, proto, args) \ + __DECLARE_TRACE_SYSCALL(name, PARAMS(proto), PARAMS(args), \ + cpu_online(raw_smp_processor_id()), \ + PARAMS(void *__data, proto)) +#define TRACE_EVENT_FLAGS (event, flag) #define TRACE_EVENT_PERF_PERM(event, expr...)
@@ -546,6 +570,9 @@ static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p) struct, assign, print) \DECLARE_TRACE_CONDITION (name ,PARAMS (proto ), \ PARAMS(args), PARAMS(cond)) +#define TRACE_EVENT_SYSCALL (name, proto, args, struct, assign, \ + print, reg, unreg) \ + DECLARE_TRACE_SYSCALL(name, PARAMS(proto), PARAMS(args))#define TRACE_EVENT_FLAGS (event, flag)
include/trace/bpf_probe.h +3 -0
@@ -53,6 +53,9 @@ __bpf_trace_##call(void *__data, proto) \#define DECLARE_EVENT_CLASS (call, proto, args, tstruct, assign, print) \ __BPF_DECLARE_TRACE(call, PARAMS(proto), PARAMS(args)) +#undef DECLARE_EVENT_SYSCALL_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +/* * This part is compiled out, it is only here as a build time check * to make sure that if the tracepoint handling changes, the
include/trace/define_trace.h +5 -0
@@ -46,6 +46,10 @@ assign, print, reg, unreg) \ DEFINE_TRACE_FN(name, reg, unreg, PARAMS(proto), PARAMS(args)) +#undef TRACE_EVENT_SYSCALL +#define TRACE_EVENT_SYSCALL (name, proto, args, struct, assign, print, reg, unreg) \ + DEFINE_TRACE_FN(name, reg, unreg, PARAMS(proto), PARAMS(args)) +#undef TRACE_EVENT_NOP#define TRACE_EVENT_NOP (name, proto, args, struct, assign, print)
@@ -107,6 +111,7 @@#undef TRACE_EVENT#undef TRACE_EVENT_FN#undef TRACE_EVENT_FN_COND +#undef TRACE_EVENT_SYSCALL#undef TRACE_EVENT_CONDITION#undef TRACE_EVENT_NOP#undef DEFINE_EVENT_NOP
include/trace/events/syscalls.h +2 -2
@@ -15,7 +15,7 @@#ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS -TRACE_EVENT_FN(sys_enter, +TRACE_EVENT_SYSCALL(sys_enter, TP_PROTO(struct pt_regs *regs, long id),
@@ -41,7 +41,7 @@ TRACE_EVENT_FN(sys_enter, TRACE_EVENT_FLAGS(sys_enter, TRACE_EVENT_FL_CAP_ANY) -TRACE_EVENT_FN(sys_exit, +TRACE_EVENT_SYSCALL(sys_exit, TP_PROTO(struct pt_regs *regs, long ret),
include/trace/perf.h +3 -0
@@ -55,6 +55,9 @@ perf_trace_##call(void *__data, proto) \ head, __task); \ } +#undef DECLARE_EVENT_SYSCALL_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +/* * This part is compiled out, it is only here as a build time check * to make sure that if the tracepoint handling changes, the
include/trace/trace_events.h +28 -0
@@ -45,6 +45,16 @@ PARAMS(print)); \DEFINE_EVENT (name, name,PARAMS (proto),PARAMS (args)); +#undef TRACE_EVENT_SYSCALL +#define TRACE_EVENT_SYSCALL (name, proto, args, tstruct, assign, print, reg, unreg) \ + DECLARE_EVENT_SYSCALL_CLASS(name, \ + PARAMS(proto), \ + PARAMS(args), \ + PARAMS(tstruct), \ + PARAMS(assign), \ + PARAMS(print)); \ + DEFINE_EVENT(name, name, PARAMS(proto), PARAMS(args)); +#include "stages/stage1_struct_define.h" #undef DECLARE_EVENT_CLASS
@@ -57,6 +67,9 @@ \ static struct trace_event_class event_class_##name; +#undef DECLARE_EVENT_SYSCALL_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +#undef DEFINE_EVENT#define DEFINE_EVENT (template, name, proto, args) \ static struct trace_event_call __used \
@@ -117,6 +130,9 @@ tstruct; \ }; +#undef DECLARE_EVENT_SYSCALL_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +#undef DEFINE_EVENT#define DEFINE_EVENT (template, name, proto, args)
@@ -208,6 +224,9 @@ static struct trace_event_functions trace_event_type_funcs_##call = { \ .trace = trace_raw_output_##call, \ }; +#undef DECLARE_EVENT_SYSCALL_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +#undef DEFINE_EVENT_PRINT#define DEFINE_EVENT_PRINT (template, call, proto, args, print) \ static notrace enum print_line_t \
@@ -265,6 +284,9 @@ static inline notrace int trace_event_get_offsets_##call( \return __data_size; \ } +#undef DECLARE_EVENT_SYSCALL_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +#include TRACE_INCLUDE (TRACE_INCLUDE_FILE )/*
@@ -409,6 +431,9 @@ trace_event_raw_event_##call(void *__data, proto) \* fail to compile unless it too is updated. */ +#undef DECLARE_EVENT_SYSCALL_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +#undef DEFINE_EVENT#define DEFINE_EVENT (template, call, proto, args) \ static inline void ftrace_test_probe_##call(void) \
@@ -434,6 +459,9 @@ static struct trace_event_class __used __refdata event_class_##call = { \ _TRACE_PERF_INIT(call) \ }; +#undef DECLARE_EVENT_SYSCALL_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +#undef DEFINE_EVENT#define DEFINE_EVENT (template, call, proto, args) \ \
Signature · 1 line
2.39.2
Mathieu Desnoyers
reviewerPATCH v4 2/8 tracing/ftrace: disable preemption in syscall probe
In reply to Mathieu DesnoyersIn preparation for allowing system call enter/exit instrumentation to handle page faults, make sure that ftrace can handle this change by explicitly disabling preemption within the ftrace system call tracepoint probes to respect the current expectations within ftrace ring buffer code.
This change does not yet allow ftrace to take page faults per se within its probe, but allows its existing probes to adapt to the upcoming change.
---
Changes since v3:
- Do not use guard() for single line function, based on feedback from Steven.
--- include/trace/trace_events.h | 39 ++++++++++++++++++++++++++++------- kernel/trace/trace_syscalls.c | 12 +++++++++++ 2 files changed, 44 insertions(+), 7 deletions(-)
include/trace/trace_events.h +32 -7
@@ -263,6 +263,9 @@ static struct trace_event_fields trace_event_fields_##call[] = { \ tstruct \ {} }; +#undef DECLARE_EVENT_SYSCALL_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +#undef DEFINE_EVENT_PRINT#define DEFINE_EVENT_PRINT (template, name, proto, args, print)
@@ -396,11 +399,11 @@ static inline notrace int trace_event_get_offsets_##call( \#include "stages/stage6_event_callback.h" -#undef DECLARE_EVENT_CLASS -#define DECLARE_EVENT_CLASS (call, proto, args, tstruct, assign, print) \ - \ + +#undef __DECLARE_EVENT_CLASS +#define __DECLARE_EVENT_CLASS (call, proto, args, tstruct, assign, print) \ static notrace void \ -trace_event_raw_event_##call(void *__data, proto) \ +do_trace_event_raw_event_##call(void *__data, proto) \ { \ struct trace_event_file *trace_file = __data; \ struct trace_event_data_offsets_##call __maybe_unused __data_offsets;\
@@ -425,15 +428,35 @@ trace_event_raw_event_##call(void *__data, proto) \ \trace_event_buffer_commit (&fbuffer); \ } + +#undef DECLARE_EVENT_CLASS +#define DECLARE_EVENT_CLASS (call, proto, args, tstruct, assign, print) \ +__DECLARE_EVENT_CLASS(call, PARAMS(proto), PARAMS(args), PARAMS(tstruct), \ + PARAMS(assign), PARAMS(print)) \ +static notrace void \ +trace_event_raw_event_##call(void *__data, proto) \ +{ \ + do_trace_event_raw_event_##call(__data, args); \ +} + +#undef DECLARE_EVENT_SYSCALL_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS (call, proto, args, tstruct, assign, print) \ +__DECLARE_EVENT_CLASS(call, PARAMS(proto), PARAMS(args), PARAMS(tstruct), \ + PARAMS(assign), PARAMS(print)) \ +static notrace void \ +trace_event_raw_event_##call(void *__data, proto) \ +{ \ + preempt_disable_notrace(); \ + do_trace_event_raw_event_##call(__data, args); \ + preempt_enable_notrace(); \ +} +/* * The ftrace_test_probe is compiled out, it is only here as a build time check * to make sure that if the tracepoint handling changes, the ftrace probe will * fail to compile unless it too is updated. */ -#undef DECLARE_EVENT_SYSCALL_CLASS -#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS -#undef DEFINE_EVENT#define DEFINE_EVENT (template, call, proto, args) \ static inline void ftrace_test_probe_##call(void) \
@@ -443,6 +466,8 @@ static inline void ftrace_test_probe_##call(void) \#include TRACE_INCLUDE (TRACE_INCLUDE_FILE ) +#undef __DECLARE_EVENT_CLASS +#include "stages/stage7_class_define.h" #undef DECLARE_EVENT_CLASS
kernel/trace/trace_syscalls.c +12 -0
@@ -299,6 +299,12 @@ static void ftrace_syscall_enter(void *data, struct pt_regs *regs, long id)int syscall_nr;int size; +/* +* Syscall probe called with preemption enabled, but the ring +* buffer and per-cpu data require preemption to be disabled. +*/ +guard (preempt_notrace)(); + syscall_nr =trace_get_syscall_nr (current, regs);if (syscall_nr <0 || syscall_nr >= NR_syscalls)return ;
@@ -338,6 +344,12 @@ static void ftrace_syscall_exit(void *data, struct pt_regs *regs, long ret)struct trace_event_buffer fbuffer;int syscall_nr; +/* +* Syscall probe called with preemption enabled, but the ring +* buffer and per-cpu data require preemption to be disabled. +*/ +guard (preempt_notrace)(); + syscall_nr =trace_get_syscall_nr (current, regs);if (syscall_nr <0 || syscall_nr >= NR_syscalls)return ;
Signature · 1 line
2.39.2
Mathieu Desnoyers
reviewerPATCH v4 3/8 tracing/perf: disable preemption in syscall probe
In reply to Mathieu DesnoyersIn preparation for allowing system call enter/exit instrumentation to handle page faults, make sure that perf can handle this change by explicitly disabling preemption within the perf system call tracepoint probes to respect the current expectations within perf ring buffer code.
This change does not yet allow perf to take page faults per se within its probe, but allows its existing probes to adapt to the upcoming change.
---
Changes since v3:
- Do not use guard() for single line function, based on feedback from Steven.
--- include/trace/perf.h | 42 +++++++++++++++++++++++++++++++---- kernel/trace/trace_syscalls.c | 12 ++++++++++ 2 files changed, 50 insertions(+), 4 deletions(-)
include/trace/perf.h +38 -4
@@ -12,10 +12,10 @@#undef __perf_task#define __perf_task (t) (__task = (t)) -#undef DECLARE_EVENT_CLASS -#define DECLARE_EVENT_CLASS (call, proto, args, tstruct, assign, print) \ +#undef __DECLARE_EVENT_CLASS +#define __DECLARE_EVENT_CLASS (call, proto, args, tstruct, assign, print) \ static notrace void \ -perf_trace_##call(void *__data, proto) \ +do_perf_trace_##call(void *__data, proto) \ { \ struct trace_event_call *event_call = __data; \ struct trace_event_data_offsets_##call __maybe_unused __data_offsets;\
@@ -55,8 +55,39 @@ perf_trace_##call(void *__data, proto) \ head, __task); \ } +/* +* Define unused __count and __task variables to use @args to pass +* arguments to do_perf_trace_##call. This is needed because the +* macros __perf_count and __perf_task introduce the side-effect to +* store copies into those local variables. +*/ +#undef DECLARE_EVENT_CLASS +#define DECLARE_EVENT_CLASS (call, proto, args, tstruct, assign, print) \ +__DECLARE_EVENT_CLASS(call, PARAMS(proto), PARAMS(args), PARAMS(tstruct), \ + PARAMS(assign), PARAMS(print)) \ +static notrace void \ +perf_trace_##call(void *__data, proto) \ +{ \ + u64 __count __attribute__((unused)); \ + struct task_struct *__task __attribute__((unused)); \ + \ + do_perf_trace_##call(__data, args); \ +} +#undef DECLARE_EVENT_SYSCALL_CLASS -#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS (call, proto, args, tstruct, assign, print) \ +__DECLARE_EVENT_CLASS(call, PARAMS(proto), PARAMS(args), PARAMS(tstruct), \ + PARAMS(assign), PARAMS(print)) \ +static notrace void \ +perf_trace_##call(void *__data, proto) \ +{ \ + u64 __count __attribute__((unused)); \ + struct task_struct *__task __attribute__((unused)); \ + \ + preempt_disable_notrace(); \ + do_perf_trace_##call(__data, args); \ + preempt_enable_notrace(); \ +}/* * This part is compiled out, it is only here as a build time check
@@ -76,4 +107,7 @@ static inline void perf_test_probe_##call(void) \ DEFINE_EVENT(template, name, PARAMS(proto), PARAMS(args))#include TRACE_INCLUDE (TRACE_INCLUDE_FILE ) + +#undef __DECLARE_EVENT_CLASS +#endif /* CONFIG_PERF_EVENTS */
kernel/trace/trace_syscalls.c +12 -0
@@ -596,6 +596,12 @@ static void perf_syscall_enter(void *ignore, struct pt_regs *regs, long id)int rctx;int size; +/* +* Syscall probe called with preemption enabled, but the ring +* buffer and per-cpu data require preemption to be disabled. +*/ +guard (preempt_notrace)(); + syscall_nr =trace_get_syscall_nr (current, regs);if (syscall_nr <0 || syscall_nr >= NR_syscalls)return ;
@@ -698,6 +704,12 @@ static void perf_syscall_exit(void *ignore, struct pt_regs *regs, long ret)int rctx;int size; +/* +* Syscall probe called with preemption enabled, but the ring +* buffer and per-cpu data require preemption to be disabled. +*/ +guard (preempt_notrace)(); + syscall_nr =trace_get_syscall_nr (current, regs);if (syscall_nr <0 || syscall_nr >= NR_syscalls)return ;
Signature · 1 line
2.39.2
Mathieu Desnoyers
reviewerPATCH v4 4/8 tracing/bpf: disable preemption in syscall probe
In reply to Mathieu DesnoyersIn preparation for allowing system call enter/exit instrumentation to handle page faults, make sure that bpf can handle this change by explicitly disabling preemption within the bpf system call tracepoint probes to respect the current expectations within bpf tracing code.
This change does not yet allow bpf to take page faults per se within its probe, but allows its existing probes to adapt to the upcoming change.
---
Changes since v3:
- Do not use guard() for single line function, based on feedback from Steven.
--- include/trace/bpf_probe.h | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)
include/trace/bpf_probe.h +11 -1
@@ -53,8 +53,18 @@ __bpf_trace_##call(void *__data, proto) \#define DECLARE_EVENT_CLASS (call, proto, args, tstruct, assign, print) \ __BPF_DECLARE_TRACE(call, PARAMS(proto), PARAMS(args)) +#define __BPF_DECLARE_TRACE_SYSCALL (call, proto, args) \ +static notrace void \ +__bpf_trace_##call(void *__data, proto) \ +{ \ + preempt_disable_notrace(); \ + CONCATENATE(bpf_trace_run, COUNT_ARGS(args))(__data, CAST_TO_U64(args)); \ + preempt_enable_notrace(); \ +} +#undef DECLARE_EVENT_SYSCALL_CLASS -#define DECLARE_EVENT_SYSCALL_CLASS DECLARE_EVENT_CLASS +#define DECLARE_EVENT_SYSCALL_CLASS (call, proto, args, tstruct, assign, print) \ + __BPF_DECLARE_TRACE_SYSCALL(call, PARAMS(proto), PARAMS(args))/* * This part is compiled out, it is only here as a build time check
Signature · 1 line
2.39.2
Mathieu Desnoyers
reviewerPATCH v4 5/8 tracing: Allow system call tracepoints to handle page faults
In reply to Mathieu DesnoyersUse Tasks Trace RCU to protect iteration of system call enter/exit tracepoint probes to allow those probes to handle page faults.
In preparation for this change, all tracers registering to system call enter/exit tracepoints should expect those to be called with preemption enabled.
This allows tracers to fault-in userspace system call arguments such as path strings within their probe callbacks.
---
Changes since v3:
- Add comment above __DO_TRACE() which describes the preempt off vs Tasks Trace RCU for syscall=0 vs syscall=1.
--- include/linux/tracepoint.h | 18 ++++++++++++++++-- init/Kconfig | 1 + 2 files changed, 17 insertions(+), 2 deletions(-)
include/linux/tracepoint.h +16 -2
@@ -17,6 +17,7 @@#include <linux/errno.h> #include <linux/types.h> #include <linux/rcupdate.h> +#include <linux/rcupdate_trace.h> #include <linux/tracepoint-defs.h> #include <linux/static_call.h>
@@ -107,6 +108,7 @@ void for_each_tracepoint_in_module(struct module *mod,#ifdef CONFIG_TRACEPOINTS static inline void tracepoint_synchronize_unregister (void ) { +synchronize_rcu_tasks_trace ();synchronize_rcu (); }#else
@@ -196,6 +198,12 @@ static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p)/* * it_func[0] is never NULL because there is at least one element in the array * when the array itself is non NULL. +* +* With @syscall=0, the tracepoint callback array dereference is +* protected by disabling preemption. +* With @syscall=1, the tracepoint callback array dereference is +* protected by Tasks Trace RCU, which allows probes to handle page +* faults. */ #define __DO_TRACE (name, args, cond, syscall) \ do { \
@@ -204,11 +212,17 @@ static inline struct tracepoint *tracepoint_ptr_deref(tracepoint_ptr_t *p)if (!(cond)) \return ; \ \ -preempt_disable_notrace (); \ +if (syscall) \ +rcu_read_lock_trace (); \ +else \ +preempt_disable_notrace (); \ \__DO_TRACE_CALL (name,TP_ARGS (args)); \ \ -preempt_enable_notrace (); \ +if (syscall) \ +rcu_read_unlock_trace (); \ +else \ +preempt_enable_notrace (); \ }while (0 )/*
init/Kconfig +1 -0
@@ -1985,6 +1985,7 @@ config BINDGEN_VERSION_TEXT # config TRACEPOINTS bool + select TASKS_TRACE_RCU source "kernel/Kconfig.kexec"
Signature · 1 line
2.39.2
Mathieu Desnoyers
reviewerPATCH v4 6/8 tracing/ftrace: Add might_fault check to syscall probes
In reply to Mathieu DesnoyersAdd a might_fault() check to validate that the ftrace sys_enter/sys_exit probe callbacks are indeed called from a context where page faults can be handled.
--- include/trace/trace_events.h | 1 + kernel/trace/trace_syscalls.c | 2 ++ 2 files changed, 3 insertions(+)
include/trace/trace_events.h +1 -0
@@ -446,6 +446,7 @@ __DECLARE_EVENT_CLASS(call, PARAMS(proto), PARAMS(args), PARAMS(tstruct), \ static notrace void \ trace_event_raw_event_##call(void *__data, proto) \ { \ + might_fault(); \ preempt_disable_notrace(); \ do_trace_event_raw_event_##call(__data, args); \ preempt_enable_notrace(); \
kernel/trace/trace_syscalls.c +2 -0
@@ -303,6 +303,7 @@ static void ftrace_syscall_enter(void *data, struct pt_regs *regs, long id)* Syscall probe called with preemption enabled, but the ring * buffer and per-cpu data require preemption to be disabled. */ +might_fault ();guard (preempt_notrace)(); syscall_nr =trace_get_syscall_nr (current, regs);
@@ -348,6 +349,7 @@ static void ftrace_syscall_exit(void *data, struct pt_regs *regs, long ret)* Syscall probe called with preemption enabled, but the ring * buffer and per-cpu data require preemption to be disabled. */ +might_fault ();guard (preempt_notrace)(); syscall_nr =trace_get_syscall_nr (current, regs);
Signature · 1 line
2.39.2
Mathieu Desnoyers
reviewerPATCH v4 7/8 tracing/perf: Add might_fault check to syscall probes
In reply to Mathieu DesnoyersAdd a might_fault() check to validate that the perf sys_enter/sys_exit probe callbacks are indeed called from a context where page faults can be handled.
--- include/trace/perf.h | 1 + kernel/trace/trace_syscalls.c | 2 ++ 2 files changed, 3 insertions(+)
include/trace/perf.h +1 -0
@@ -84,6 +84,7 @@ perf_trace_##call(void *__data, proto) \ u64 __count __attribute__((unused)); \ struct task_struct *__task __attribute__((unused)); \ \ +might_fault (); \preempt_disable_notrace (); \ do_perf_trace_##call(__data, args); \ preempt_enable_notrace(); \
kernel/trace/trace_syscalls.c +2 -0
@@ -602,6 +602,7 @@ static void perf_syscall_enter(void *ignore, struct pt_regs *regs, long id)* Syscall probe called with preemption enabled, but the ring * buffer and per-cpu data require preemption to be disabled. */ +might_fault ();guard (preempt_notrace)(); syscall_nr =trace_get_syscall_nr (current, regs);
@@ -710,6 +711,7 @@ static void perf_syscall_exit(void *ignore, struct pt_regs *regs, long ret)* Syscall probe called with preemption enabled, but the ring * buffer and per-cpu data require preemption to be disabled. */ +might_fault ();guard (preempt_notrace)(); syscall_nr =trace_get_syscall_nr (current, regs);
Signature · 1 line
2.39.2
Mathieu Desnoyers
reviewerPATCH v4 8/8 tracing/bpf: Add might_fault check to syscall probes
In reply to Mathieu DesnoyersAdd a might_fault() check to validate that the bpf sys_enter/sys_exit probe callbacks are indeed called from a context where page faults can be handled.
--- include/trace/bpf_probe.h | 1 + 1 file changed, 1 insertion(+)
include/trace/bpf_probe.h +1 -0
@@ -57,6 +57,7 @@ __bpf_trace_##call(void *__data, proto) \ static notrace void \ __bpf_trace_##call(void *__data, proto) \ { \ + might_fault(); \ preempt_disable_notrace(); \ CONCATENATE(bpf_trace_run, COUNT_ARGS(args))(__data, CAST_TO_U64(args)); \ preempt_enable_notrace(); \
Signature · 1 line
2.39.2