tracing/perf: Use helper functions to help shrink kernel size
Steven Rostedt
maintainercover tracing/perf: Use helper functions to help shrink kernel size
This patch set is another effort to bring down the size of trace events.
Helper functions are used to remove duplicate code that is created for each event by the TRACE_EVENT() macros. With a config that enables 707 events in the core kernel, the size is brought down 91,774 bytes!
Steven Rostedt (4):
tracing: Move print code from macro to standalone function
tracing: Move event storage for array from macro to standalone function
tracing: Use helper functions in event assignment to shrink macro size perf/events: Use helper functions in event assignment to shrink macro size
----
include/linux/ftrace_event.h | 40 +++++++++++++++-- include/trace/ftrace.h | 85 +++++++++--------------------------- kernel/trace/trace_event_perf.c | 26 +++++++++++ kernel/trace/trace_events.c | 6 --- kernel/trace/trace_export.c | 12 ++---- kernel/trace/trace_output.c | 90 +++++++++++++++++++++++++++++++++++++++ 6 files changed, 176 insertions(+), 83 deletions(-)
Steven Rostedt
maintainerRFC PATCH 1/4 tracing: Move print code from macro to standalone function
From: Steven Rostedt <[email protected]>
The code for trace events to format the raw recorded event data into human readable format in the 'trace' file is repeated for every event in the system. When you have over 700 events, this can add up quite a bit.
By making helper functions in the core kernel to do the work instead, we can shrink the size of the kernel down a bit.
With a kernel configured with 707 events, the change in size was:
text data bss dec hex filename
20016471 2594648 1945600 24556719 176b4af vmlinux-before 19994715 2594648 1945600 24534963 1765fb3 vmlinux-after
That's a total of 21756 bytes, which comes down to 30 bytes per event.
--- include/linux/ftrace_event.h | 5 +++++ include/trace/ftrace.h | 21 ++------------------ kernel/trace/trace_output.c | 44 ++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 51 insertions(+), 19 deletions(-)
include/linux/ftrace_event.h +5 -0
@@ -139,6 +139,11 @@ void trace_current_buffer_discard_commit(struct ring_buffer *buffer,void tracing_record_cmdline (struct task_struct *tsk); +int ftrace_output_event (struct trace_iterator *iter,struct ftrace_event_call *event, +char *fmt, ...); + +int ftrace_output_call (struct trace_iterator *iter,char *name,char *fmt, ...); +struct event_filter ;enum trace_reg {
include/trace/ftrace.h +2 -19
@@ -228,11 +228,9 @@ ftrace_raw_output_##call(struct trace_iterator *iter, int flags, \ struct trace_event *trace_event) \ { \struct ftrace_event_call *event; \ -struct trace_seq *s = &iter->seq; \ struct ftrace_raw_##call *field; \ struct trace_entry *entry; \ struct trace_seq *p = &iter->tmp_seq; \ - int ret; \ \ event = container_of(trace_event, struct ftrace_event_call, \ event); \
@@ -245,15 +243,8 @@ ftrace_raw_output_##call(struct trace_iterator *iter, int flags, \ } \ \ field = (typeof (field ))entry; \ - \trace_seq_init (p); \ - ret =trace_seq_printf (s,"%s: " , event->name); \ -if (ret) \ - ret =trace_seq_printf (s, print); \ -if (!ret) \ -return TRACE_TYPE_PARTIAL_LINE ; \ - \ -return TRACE_TYPE_HANDLED ; \ +return ftrace_output_event (iter, event, print); \ } \ static struct trace_event_functions ftrace_event_type_funcs_##call = { \ .trace = ftrace_raw_output_##call, \
@@ -265,11 +256,9 @@ static notrace enum print_line_t \ ftrace_raw_output_##call(struct trace_iterator *iter, int flags, \ struct trace_event *event) \ { \ - struct trace_seq *s = &iter->seq; \ struct ftrace_raw_##template *field; \ struct trace_entry *entry; \ struct trace_seq *p = &iter->tmp_seq; \ - int ret; \ \ entry = iter->ent; \ \
@@ -281,13 +270,7 @@ ftrace_raw_output_##call(struct trace_iterator *iter, int flags, \ field = (typeof (field ))entry; \ \trace_seq_init (p); \ - ret = trace_seq_printf(s, "%s: ", #call); \ - if (ret) \ - ret = trace_seq_printf(s, print); \ - if (!ret) \ - return TRACE_TYPE_PARTIAL_LINE; \ - \ - return TRACE_TYPE_HANDLED; \ + return ftrace_output_call(iter, #call, print); \ } \ static struct trace_event_functions ftrace_event_type_funcs_##call = { \ .trace = ftrace_raw_output_##call, \
kernel/trace/trace_output.c +44 -0
50 lines not on this page · raw message
Signature · 1 line
1.7.10.4
Steven Rostedt
maintainerRFC PATCH 2/4 tracing: Move event storage for array from macro to standalone function
In reply to Steven RostedtFrom: Steven Rostedt <[email protected]>
The code that shows array fields for events is defined for all events. This can add up quite a bit when you have over 700 events.
By making helper functions in the core kernel to do the work instead, we can shrink the size of the kernel down a little.
With a kernel configured with 707 events, the change in size was:
text data bss dec hex filename
19994715 2594648 1945600 24534963 1765fb3 vmlinux-before 19991705 2594648 1945600 24531953 17653f1 vmlinux-after
That's a total of 3010 bytes, which comes down to 4 bytes per event. Although it's not much, this code is just called at initialization.
--- include/linux/ftrace_event.h | 8 ++++---- include/trace/ftrace.h | 12 ++++-------- kernel/trace/trace_events.c | 6 ------ kernel/trace/trace_export.c | 12 ++++-------- kernel/trace/trace_output.c | 20 ++++++++++++++++++++ 5 files changed, 32 insertions(+), 26 deletions(-)
include/linux/ftrace_event.h +4 -4
@@ -144,6 +144,10 @@ int ftrace_output_event(struct trace_iterator *iter, struct ftrace_event_call *eint ftrace_output_call (struct trace_iterator *iter,char *name,char *fmt, ...); +int ftrace_event_define_field (struct ftrace_event_call *call, +char *type,int len,char *item,int offset, +int field_size,int sign,int filter); +struct event_filter ;enum trace_reg {
@@ -260,10 +264,6 @@ enum { FILTER_TRACE_FN, }; -#define EVENT_STORAGE_SIZE 128 -extern struct mutex event_storage_mutex; -extern char event_storage[EVENT_STORAGE_SIZE ]; -extern int trace_event_raw_init (struct ftrace_event_call *call); extern int trace_define_field(struct ftrace_event_call *call, const char *type, const char *name, int offset, int size,
include/trace/ftrace.h +4 -8
@@ -293,15 +293,11 @@ static struct trace_event_functions ftrace_event_type_funcs_##call = { \#undef __array#define __array (type, item, len) \ do { \ - mutex_lock(&event_storage_mutex); \ BUILD_BUG_ON(len > MAX_FILTER_STR_VAL); \ - snprintf(event_storage, sizeof(event_storage), \ - "%s[%d]", #type, len); \ - ret = trace_define_field(event_call, event_storage, #item, \ - offsetof(typeof(field), item), \ - sizeof(field.item), \ - is_signed_type(type), FILTER_OTHER); \ - mutex_unlock(&event_storage_mutex); \ + ret = ftrace_event_define_field(event_call, #type, len, \ + #item, offsetof(typeof(field), item), \ + sizeof(field.item), \ + is_signed_type(type), FILTER_OTHER); \ if (ret) \ return ret; \ } while (0);
kernel/trace/trace_events.c +0 -6
@@ -27,12 +27,6 @@DEFINE_MUTEX (event_mutex); -DEFINE_MUTEX (event_storage_mutex); -EXPORT_SYMBOL_GPL (event_storage_mutex); - -char event_storage[EVENT_STORAGE_SIZE ]; -EXPORT_SYMBOL_GPL (event_storage); -LIST_HEAD (ftrace_events);LIST_HEAD (ftrace_common_fields);
kernel/trace/trace_export.c +4 -8
@@ -96,14 +96,10 @@ static void __always_unused ____ftrace_check_##name(void) \#define __array (type, item, len) \ do { \ BUILD_BUG_ON(len > MAX_FILTER_STR_VAL); \ - mutex_lock(&event_storage_mutex); \ - snprintf(event_storage, sizeof(event_storage), \ - "%s[%d]", #type, len); \ - ret = trace_define_field(event_call, event_storage, #item, \ - offsetof(typeof(field), item), \ - sizeof(field.item), \ - is_signed_type(type), filter_type); \ - mutex_unlock(&event_storage_mutex); \ + ret = ftrace_event_define_field(event_call, #type, len, \ + #item, offsetof(typeof(field), item), \ + sizeof(field.item), \ + is_signed_type(type), filter_type); \ if (ret) \ return ret; \ } while (0);
kernel/trace/trace_output.c +20 -0
@@ -20,6 +20,10 @@ static struct hlist_head event_hash[EVENT_HASHSIZE] __read_mostly;static int next_event_type = __TRACE_LAST_TYPE +1 ; +#define EVENT_STORAGE_SIZE 128 +static DEFINE_MUTEX (event_storage_mutex); +static char event_storage[EVENT_STORAGE_SIZE ]; +int trace_print_seq (struct seq_file *m,struct trace_seq *s) {int len = s->len >=PAGE_SIZE ?PAGE_SIZE -1 : s->len;
@@ -441,6 +445,22 @@ int ftrace_output_event(struct trace_iterator *iter, struct ftrace_event_call *e }EXPORT_SYMBOL_GPL (ftrace_output_event); +int ftrace_event_define_field (struct ftrace_event_call *call, +char *type,int len,char *item,int offset, +int field_size,int sign,int filter) +{ +int ret; + +mutex_lock (&event_storage_mutex); +snprintf (event_storage,sizeof (event_storage), +"%s[%d]" , type, len); + ret =trace_define_field (call, event_storage, item, offset, + field_size, sign, filter); +mutex_unlock (&event_storage_mutex); + +return ret; +} +#ifdef CONFIG_KRETPROBES static inline const char *kretprobed (const char *name) {
Signature · 1 line
1.7.10.4
Steven Rostedt
maintainerRFC PATCH 3/4 tracing: Use helper functions in event assignment to shrink macro size
In reply to Steven RostedtFrom: Steven Rostedt <[email protected]>
The functions that assign the contents for the ftrace events are defined by the TRACE_EVENT() macros. Each event has its own unique way to assign data to its buffer. When you have over 700 events, that means there's 700 functions assigning data uniquely for each event (not really that many, as DECLARE_EVENT_CLASS() and multiple DEFINE_EVENT()s will only need a single function).
By making helper functions in the core kernel to do some of the work instead, we can shrink the size of the kernel down a bit.
With a kernel configured with 707 events, the change in size was:
text data bss dec hex filename
19991705 2594648 1945600 24531953 17653f1 vmlinux-before 19966937 2594648 1945600 24507185 175f331 vmlinux-after
That's a total of 24768 bytes, which comes down to 35 bytes per event.
--- include/linux/ftrace_event.h | 14 ++++++++++++++ include/trace/ftrace.h | 21 ++++++--------------- kernel/trace/trace_output.c | 26 ++++++++++++++++++++++++++ 3 files changed, 46 insertions(+), 15 deletions(-)
include/linux/ftrace_event.h +14 -0
@@ -148,6 +148,20 @@ int ftrace_event_define_field(struct ftrace_event_call *call, char *type, int len, char *item, int offset, int field_size, int sign, int filter); +struct ftrace_event_buffer { +struct ring_buffer *buffer; +struct ring_buffer_event *event; +unsigned long flags; +int pc; +}; + +void *ftrace_event_buffer_reserve (struct ftrace_event_buffer *fbuffer, +int type,unsigned long len); + +void ftrace_event_buffer_commit (struct ftrace_event_buffer *fbuffer, +struct ftrace_event_call *event_call, +void *entry); +struct event_filter ;enum trace_reg {
include/trace/ftrace.h +6 -15
@@ -499,33 +499,24 @@ ftrace_raw_event_##call(void *__data, proto) \ { \struct ftrace_event_call *event_call = __data; \ struct ftrace_data_offsets_##call __maybe_unused __data_offsets;\ - struct ring_buffer_event *event; \ + struct ftrace_event_buffer fbuffer; \ struct ftrace_raw_##call *entry; \ - struct ring_buffer *buffer; \ - unsigned long irq_flags; \ int __data_size; \ - int pc; \ - \ - local_save_flags(irq_flags); \ - pc = preempt_count(); \ \ __data_size = ftrace_get_offsets_##call(&__data_offsets, args); \ \ - event = trace_current_buffer_lock_reserve(&buffer, \ + entry = ftrace_event_buffer_reserve(&fbuffer, \ event_call->event.type, \ - sizeof(*entry) + __data_size, \ - irq_flags, pc); \ - if (!event) \ + sizeof(*entry) + __data_size); \ + \ + if (!entry) \ return; \ - entry = ring_buffer_event_data(event); \ \ tstruct \ \ { assign; } \ \ - if (!filter_current_check_discard(buffer, event_call, entry, event)) \ - trace_nowake_buffer_unlock_commit(buffer, \ - event, irq_flags, pc); \ + ftrace_event_buffer_commit(&fbuffer, event_call, entry); \ }/* * The ftrace_test_probe is compiled out, it is only here as a build time check
kernel/trace/trace_output.c +26 -0
@@ -461,6 +461,32 @@ int ftrace_event_define_field(struct ftrace_event_call *call,return ret; } +void *ftrace_event_buffer_reserve (struct ftrace_event_buffer *fbuffer, +int type,unsigned long len) +{ +local_save_flags (fbuffer->flags); + fbuffer->pc =preempt_count (); + + fbuffer->event = +trace_current_buffer_lock_reserve (&fbuffer->buffer, type, len, + fbuffer->flags, fbuffer->pc); +if (!fbuffer->event) +return NULL ; + +return ring_buffer_event_data (fbuffer->event); +} + +void ftrace_event_buffer_commit (struct ftrace_event_buffer *fbuffer, +struct ftrace_event_call *event_call, +void *entry) +{ +if (!filter_current_check_discard (fbuffer->buffer, event_call, + entry, fbuffer->event)) +trace_nowake_buffer_unlock_commit (fbuffer->buffer, + fbuffer->event, + fbuffer->flags, fbuffer->pc); +} +#ifdef CONFIG_KRETPROBES static inline const char *kretprobed (const char *name) {
Signature · 1 line
1.7.10.4
Steven Rostedt
maintainerRFC PATCH 4/4 perf/events: Use helper functions in event assignment to shrink macro size
In reply to Steven RostedtFrom: Steven Rostedt <[email protected]>
The functions that assign the contents for the perf software events are defined by the TRACE_EVENT() macros. Each event has its own unique way to assign data to its buffer. When you have over 700 events, that means there's 700 functions assigning data uniquely for each event.
By making helper functions in the core kernel to do the work instead, we can shrink the size of the kernel down a bit.
With a kernel configured with 707 events, the change in size was:
text data bss dec hex filename
19966937 2594648 1945600 24507185 175f331 vmlinux-before 19924761 2594584 1945600 24464945 1754e31 vmlinux-after
That's a total of 42240 bytes, which comes down to 59 bytes per event.
--- include/linux/ftrace_event.h | 13 +++++++++++++ include/trace/ftrace.h | 31 ++++++++----------------------- kernel/trace/trace_event_perf.c | 26 ++++++++++++++++++++++++++ 3 files changed, 47 insertions(+), 23 deletions(-)
include/linux/ftrace_event.h +13 -0
@@ -313,6 +313,19 @@ struct perf_event; DECLARE_PER_CPU(struct pt_regs, perf_trace_regs); +struct perf_trace_event { +struct pt_regs regs; +u64 addr; +u64 count; +int entry_size; +int rctx; +}; + +extern void *perf_trace_event_setup (struct ftrace_event_call *event_call, +struct perf_trace_event *pe); +extern void perf_trace_event_submit (void *raw_data,struct ftrace_event_call *event_call, +struct perf_trace_event *pe); +extern int perf_trace_init (struct perf_event *event);extern void perf_trace_destroy (struct perf_event *event);extern int perf_trace_add (struct perf_event *event,int flags);
include/trace/ftrace.h +8 -23
@@ -677,10 +677,10 @@ __attribute__((section("_ftrace_events"))) *__event_##call = &event_##call#define __get_str (field) (char *)__get_dynamic_array(field)#undef __perf_addr -#define __perf_addr (a) __addr = (a) +#define __perf_addr (a) __pe.addr = (a)#undef __perf_count -#define __perf_count (c) __count = (c) +#define __perf_count (c) __pe.count = (c)#undef TP_perf_assign #define TP_perf_assign(args...) args
@@ -693,36 +693,21 @@ perf_trace_##call(void *__data, proto) \struct ftrace_event_call *event_call = __data; \ struct ftrace_data_offsets_##call __maybe_unused __data_offsets;\ struct ftrace_raw_##call *entry; \ - struct pt_regs __regs; \ - u64 __addr = 0, __count = 1; \ - struct hlist_head *head; \ - int __entry_size; \ + struct perf_trace_event __pe; \ int __data_size; \ - int rctx; \ - \ - perf_fetch_caller_regs(&__regs); \ \ __data_size = ftrace_get_offsets_##call(&__data_offsets, args); \ - __entry_size = ALIGN(__data_size + sizeof(*entry) + sizeof(u32),\ - sizeof(u64)); \ - __entry_size -= sizeof(u32); \ - \ - if (WARN_ONCE(__entry_size > PERF_MAX_TRACE_SIZE, \ - "profile buffer not large enough")) \ - return; \ + __pe.entry_size = __data_size + sizeof(*entry); \ + __pe.addr = 0; \ + __pe.count = 1; \ \ - entry = (struct ftrace_raw_##call *)perf_trace_buf_prepare( \ - __entry_size, event_call->event.type, &__regs, &rctx); \ - if (!entry) \ - return; \ + entry = perf_trace_event_setup(event_call, &__pe); \ \ tstruct \ \ { assign; } \ \ - head = this_cpu_ptr(event_call->perf_events); \ - perf_trace_buf_submit(entry, __entry_size, rctx, __addr, \ - __count, &__regs, head); \ + perf_trace_event_submit(entry, event_call, &__pe); \ }/*
kernel/trace/trace_event_perf.c +26 -0
@@ -21,6 +21,32 @@ typedef typeof(unsigned long [PERF_MAX_TRACE_SIZE / sizeof(unsigned long)])/* Count the events in use (per event id, not per instance) */ static int total_ref_count; +void *perf_trace_event_setup (struct ftrace_event_call *event_call, +struct perf_trace_event *pe) +{ +perf_fetch_caller_regs (&pe->regs); + + pe->entry_size =ALIGN (pe->entry_size +sizeof (u32),sizeof (u64)); + pe->entry_size -=sizeof (u32); + +if (WARN_ONCE (pe->entry_size >PERF_MAX_TRACE_SIZE , +"profile buffer not large enough" )) +return NULL ; + +return perf_trace_buf_prepare (pe->entry_size, + event_call->event.type, &pe->regs, &pe->rctx); +} + +void perf_trace_event_submit (void *raw_data,struct ftrace_event_call *event_call, +struct perf_trace_event *pe) +{ +struct hlist_head *head; + + head =this_cpu_ptr (event_call->perf_events); +perf_trace_buf_submit (raw_data, pe->entry_size, pe->rctx, pe->addr, + pe->count, &pe->regs, head); +} +static int perf_trace_event_perm (struct ftrace_event_call *tp_event,struct perf_event *p_event) {
Signature · 1 line
1.7.10.4
Re: [RFC][PATCH 4/4] perf/events: Use helper functions in event assignment to shrink macro size
Quoted 13 lines · Steven Rostedt
plain text document attachment (0004-perf-events-Use-helper-functions-in-event-assignment.patch)
From: Steven Rostedt <[email protected]>The functions that assign the contents for the perf software events are defined by the
TRACE_EVENT()macros. Each event has its own unique way to assign data to its buffer. When you have over 700 events, that means there's 700 functions assigning data uniquely for each event.By making helper functions in the core kernel to do the work instead, we can shrink the size of the kernel down a bit.
Quoted 7 lines
With a kernel configured with 707 events, the change in size was:
text data bss dec hex filename
19966937 2594648 1945600 24507185 175f331 vmlinux-before 19924761 2594584 1945600 24464945 1754e31 vmlinux-afterThat's a total of 42240 bytes, which comes down to 59 bytes per event.
Seems like good stuff.. one nit..
kernel/trace/trace_event_perf.c typedef typeof(unsigned long [PERF_MAX_TRACE_SIZE / sizeof(unsigned long)]) · line 48
Quoted 9 lines
kernel/trace/trace_event_perf.c +9
-0+void perf_trace_event_submit (void *raw_data,struct ftrace_event_call *event_call, +struct perf_trace_event *pe) +{ +struct hlist_head *head; + + head =this_cpu_ptr (event_call->perf_events); +perf_trace_buf_submit (raw_data, pe->entry_size, pe->rctx, pe->addr, + pe->count, &pe->regs, head); +}
Can you make perf_trace_buf_submit() go away? Its reduced to a simple fwd function and layering another wrapper on top seems like pushing it.
Steven Rostedt
maintainerRe: [RFC][PATCH 4/4] perf/events: Use helper functions in event assignment to shrink macro size
Quoted 4 lines · Peter Zijlstra
Quoted 4 lines
kernel/trace/trace_event_perf.c +4
-0+void perf_trace_event_submit (void *raw_data,struct ftrace_event_call *event_call, +struct perf_trace_event *pe) +{ +struct hlist_head *head;
Quoted 8 lines
kernel/trace/trace_event_perf.c typedef typeof(unsigned long [PERF_MAX_TRACE_SIZE / sizeof(unsigned long)]) · line 48
Quoted 5 lines
kernel/trace/trace_event_perf.c +5
-0+ + head =this_cpu_ptr (event_call->perf_events); +perf_trace_buf_submit (raw_data, pe->entry_size, pe->rctx, pe->addr, + pe->count, &pe->regs, head); +}Can you make
perf_trace_buf_submit()go away? Its reduced to a simple fwd function and layering another wrapper on top seems like pushing it.
You mean just have perf_trace_event_submit() call perf_tp_event() directly?
I have no problem with that. Although I may make that into a separate patch to keep this patch as a 'move' and the other patch as the change.
Looking at the history of perf_trace_buf_submit(), it use to be more than one function call. But when you inlined perf_swevent_put_recursion_context(), it became just a one2one mapping.
I'm assuming that we want to convert all calls to
perf_trace_buf_submit()s into perf_tp_event()?
-- Steve
Re: [RFC][PATCH 4/4] perf/events: Use helper functions in event assignment to shrink macro size
Quoted 18 lines · Steven Rostedt
Quoted 4 lines · Peter Zijlstra
Quoted 4 lines
kernel/trace/trace_event_perf.c +4
-0+void perf_trace_event_submit (void *raw_data,struct ftrace_event_call *event_call, +struct perf_trace_event *pe) +{ +struct hlist_head *head;Quoted 8 lines
kernel/trace/trace_event_perf.c typedef typeof(unsigned long [PERF_MAX_TRACE_SIZE / sizeof(unsigned long)]) · line 48
Quoted 5 lines
kernel/trace/trace_event_perf.c +5
-0+ + head =this_cpu_ptr (event_call->perf_events); +perf_trace_buf_submit (raw_data, pe->entry_size, pe->rctx, pe->addr, + pe->count, &pe->regs, head); +}Can you make
perf_trace_buf_submit()go away? Its reduced to a simple fwd function and layering another wrapper on top seems like pushing it.You mean just have
perf_trace_event_submit()callperf_tp_event()directly?
Quoted 6 lines
I have no problem with that. Although I may make that into a separate patch to keep this patch as a 'move' and the other patch as the change.
Looking at the history of
perf_trace_buf_submit(), it use to be more than one function call. But when you inlinedperf_swevent_put_recursion_context(), it became just a one2one mapping.
Right.
Quoted 2 lines
I'm assuming that we want to convert all calls to
perf_trace_buf_submit()s intoperf_tp_event()?
Yeah.. I think you're referring to the {u,k}probes open-coded nonsense? Should we make those use these new helpers you created as well?
Steven Rostedt
maintainerRe: [RFC][PATCH 4/4] perf/events: Use helper functions in event assignment to shrink macro size
Quoted 5 lines · Peter Zijlstra
Quoted 3 lines
Looking at the history of
perf_trace_buf_submit(), it use to be more than one function call. But when you inlinedperf_swevent_put_recursion_context(), it became just a one2one mapping.Right.
OK, this should go as a separate change.
Quoted 6 lines
Quoted 3 lines
I'm assuming that we want to convert all calls to
perf_trace_buf_submit()s intoperf_tp_event()?Yeah.. I think you're referring to the {u,k}probes open-coded nonsense? Should we make those use these new helpers you created as well?
They may not need it. The helpers I created were for the macros that are created by all TRACE_EVENT() instances. If the uprobes and kprobes only have a single instance, then they probably don't need the helpers. But they still call the perf_trace_buf_submit(), which can can probably be cleaned up.
-- Steve
Re: [RFC][PATCH 4/4] perf/events: Use helper functions in event assignment to shrink macro size
Quoted 8 lines · Steven Rostedt
Quoted 2 lines
Yeah.. I think you're referring to the {u,k}probes open-coded nonsense? Should we make those use these new helpers you created as well?
They may not need it. The helpers I created were for the macros that are created by all
TRACE_EVENT()instances. If the uprobes and kprobes only have a single instance, then they probably don't need the helpers. But they still call theperf_trace_buf_submit(), which can can probably be cleaned up.
Right, I realize that. I was just thinking it might be good to keep all (3?) copies similar. Also avoiding two sets of wrappers seems like a good idea.