perf tools: Factorize common code
Ingo,
This set factorizes the common definition of structs, functions, variables inside perf tools to make it easier for perf trace to be created.
I guess this patchset only factorizes 10 % of what could be done, the largest part is the intersection between perf report and perf annotate. It's a lot of work.
I guess I will do the rest in separate patchsets while writing perf trace.
Thanks,
Frederic.
The following changes since commit 9f8666971185b86615a074bcac67c90fdf8af8bc:
Brice Goglin (1):
perf report: Add raw displaying of per-thread counters
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing.git ..BRANCH.NOT.VERIFIED..
Frederic Weisbecker (3):
perf tools: Factorize high level dso helpers
perf tools: Factorize the event structure definitions in a single file perf tools: Factorize the map helpers
tools/perf/Makefile | 2 + tools/perf/builtin-annotate.c | 209 +------------------------------- tools/perf/builtin-record.c | 20 +--- tools/perf/builtin-report.c | 274 +---------------------------------------- tools/perf/builtin-stat.c | 1 - tools/perf/builtin-top.c | 24 ---- tools/perf/builtin.h | 1 + tools/perf/perf.h | 1 + tools/perf/util/callchain.h | 1 + tools/perf/util/event.h | 84 +++++++++++++ tools/perf/util/map.c | 97 +++++++++++++++ tools/perf/util/symbol.c | 97 +++++++++++++++ tools/perf/util/symbol.h | 12 ++ tools/perf/util/util.h | 1 + 14 files changed, 303 insertions(+), 521 deletions(-) create mode 100644 tools/perf/util/event.h create mode 100644 tools/perf/util/map.c
PATCH 1/3 perf tools: Factorize high level dso helpers
Factorize multiple definitions of high level dso helpers into the symbol source file.
The side effect is a general export of the verbose and eprintf debugging helpers into a new file dedicated to debugging purposes.
--- tools/perf/Makefile | 1 + tools/perf/builtin-annotate.c | 96 ---------------------------------------- tools/perf/builtin-record.c | 1 - tools/perf/builtin-report.c | 97 ----------------------------------------- tools/perf/builtin-stat.c | 1 - tools/perf/builtin-top.c | 4 -- tools/perf/builtin.h | 1 + tools/perf/perf.h | 1 + tools/perf/util/symbol.c | 97 +++++++++++++++++++++++++++++++++++++++++ tools/perf/util/symbol.h | 11 +++++ 10 files changed, 111 insertions(+), 199 deletions(-)
tools/perf/Makefile +1 -0
@@ -339,6 +339,7 @@ LIB_OBJS += util/pager.o LIB_OBJS += util/header.o LIB_OBJS += util/callchain.o LIB_OBJS += util/values.o +LIB_OBJS += util/debug.o BUILTIN_OBJS += builtin-annotate.o BUILTIN_OBJS += builtin-help.o
tools/perf/builtin-annotate.c +0 -96
114 lines not on this page · raw message
tools/perf/builtin-record.c +0 -1
@@ -40,7 +40,6 @@ static int inherit = 1;static int force =0 ;static int append_file =0 ;static int call_graph =0 ; -static int verbose =0 ;static int inherit_stat =0 ;static int no_samples =0 ;static int sample_address =0 ;
tools/perf/builtin-report.c +0 -97
116 lines not on this page · raw message
tools/perf/builtin-stat.c +0 -1
@@ -63,7 +63,6 @@ static struct perf_counter_attr default_attrs[] = {#define MAX_RUN 100static int system_wide =0 ; -static int verbose =0 ;static unsigned int nr_cpus =0 ;static int run_idx =0 ;
tools/perf/builtin-top.c +0 -4
@@ -68,8 +68,6 @@ static int group = 0;static unsigned int page_size;static unsigned int mmap_pages =16 ;static int freq =0 ; -static int verbose =0 ; -static char *vmlinux =NULL ;static int delay_secs =2 ;static int zero;
@@ -338,8 +336,6 @@ static void show_details(struct sym_entry *syme)printf ("%d lines not displayed, maybe increase display entries [e]\n" , more); } -struct dso *kernel_dso; -/* * Symbols will be added here in record_ip and will get out * after decayed.
tools/perf/builtin.h +1 -0
@@ -22,5 +22,6 @@ extern int cmd_stat(int argc, const char **argv, const char *prefix);extern int cmd_top (int argc,const char **argv,const char *prefix);extern int cmd_version (int argc,const char **argv,const char *prefix);extern int cmd_list (int argc,const char **argv,const char *prefix); +extern int cmd_trace (int argc,const char **argv,const char *prefix);#endif
tools/perf/perf.h +1 -0
@@ -48,6 +48,7 @@#include "../../include/linux/perf_counter.h" #include "util/types.h" +#include "util/debug.h" /* * prctl(PR_TASK_PERF_COUNTERS_DISABLE) will (cheaply) disable all
tools/perf/util/symbol.c +97 -0
103 lines not on this page · raw message
tools/perf/util/symbol.h +11 -0
@@ -48,9 +48,20 @@ int dso__load_kernel(struct dso *self, const char *vmlinux, symbol_filter_t filter, int verbose, int modules);int dso__load_modules (struct dso *self,symbol_filter_t filter,int verbose);int dso__load (struct dso *self,symbol_filter_t filter,int verbose); +struct dso *dsos__findnew (const char *name); +void dsos__fprintf (FILE *fp);size_t dso__fprintf (struct dso *self,FILE *fp);char dso__symtab_origin (const struct dso *self); +int load_kernel (void ); +void symbol__init (void ); + +extern struct list_head dsos; +extern struct dso *kernel_dso; +extern struct dso *vdso; +extern struct dso *hypervisor_dso; +extern char *vmlinux; +extern int modules;#endif /* _PERF_SYMBOL_ */
Signature · 1 line
1.6.2.3
Re: [PATCH 1/3] perf tools: Factorize high level dso helpers
Quoted 29 lines · Frederic Weisbecker
Factorize multiple definitions of high level dso helpers into the symbol source file.
The side effect is a general export of the verbose and eprintf debugging helpers into a new file dedicated to debugging purposes.
Signed-off-by Frederic Weisbecker <[email protected]>Cc Arnaldo Carvalho de Melo <[email protected]>Cc Peter Zijlstra <[email protected]>Cc Mike Galbraith <[email protected]>Cc Brice Goglin <[email protected]>--- tools/perf/Makefile | 1 + tools/perf/builtin-annotate.c | 96 ---------------------------------------- tools/perf/builtin-record.c | 1 - tools/perf/builtin-report.c | 97 ----------------------------------------- tools/perf/builtin-stat.c | 1 - tools/perf/builtin-top.c | 4 -- tools/perf/builtin.h | 1 + tools/perf/perf.h | 1 + tools/perf/util/symbol.c | 97 +++++++++++++++++++++++++++++++++++++++++ tools/perf/util/symbol.h | 11 +++++ 10 files changed, 111 insertions(+), 199 deletions(-)tools/perf/Makefile +0
-0
tools/perf/Makefile LIB_OBJS += util/pager.o · line 342
Quoted 4 lines
tools/perf/Makefile +1
-0@@ -339,6 +339,7 @@ LIB_OBJS += util/pager.o LIB_OBJS += util/header.o LIB_OBJS += util/callchain.o LIB_OBJS += util/values.o +LIB_OBJS += util/debug.o
Damn, did I forgot to git-add util/debug.c ?
How could I fix this? Another pull request?
Frederic.
Quoted 452 lines
tools/perf/Makefile +0
-0BUILTIN_OBJS += builtin-annotate.o BUILTIN_OBJS += builtin-help.otools/perf/builtin-annotate.c +0
-96@@ -26,7 +26,6 @@#define SHOW_HV 4static char const *input_name ="perf.data" ; -static char *vmlinux ="vmlinux" ;static char default_sort_order[] ="comm,symbol" ;static char *sort_order = default_sort_order;@@ -37,9 +36,6 @@ static int show_mask = SHOW_KERNEL | SHOW_USER | SHOW_HV;static int dump_trace =0 ; #define dprintf(x...) do { if (dump_trace) printf(x); } while (0) -static int verbose; - -static int modules;static int full_paths;@@ -89,98 +85,6 @@ struct sym_ext {char *path; }; -static LIST_HEAD (dsos); -static struct dso *kernel_dso; -static struct dso *vdso; - - -static void dsos__add (struct dso *dso) -{ -list_add_tail (&dso->node, &dsos); -} - -static struct dso *dsos__find (const char *name) -{ -struct dso *pos; - -list_for_each_entry (pos, &dsos, node) -if (strcmp (pos->name, name) ==0 ) -return pos; -return NULL ; -} - -static struct dso *dsos__findnew (const char *name) -{ -struct dso *dso =dsos__find (name); -int nr; - -if (dso) -return dso; - - dso =dso__new (name,0 ); -if (!dso) - goto out_delete_dso; - - nr =dso__load (dso,NULL , verbose); -if (nr <0 ) { -if (verbose) -fprintf (stderr,"Failed to open: %s\n" , name); - goto out_delete_dso; - } -if (!nr && verbose) { -fprintf (stderr, -"No symbols found in: %s, maybe install a debug package?\n" , - name); - } - -dsos__add (dso); - -return dso; - -out_delete_dso: -dso__delete (dso); -return NULL ; -} - -static void dsos__fprintf (FILE *fp) -{ -struct dso *pos; - -list_for_each_entry (pos, &dsos, node) -dso__fprintf (pos, fp); -} - -static struct symbol *vdso__find_symbol (struct dso *dso,u64 ip) -{ -return dso__find_symbol (dso, ip); -} - -static int load_kernel (void ) -{ -int err; - - kernel_dso =dso__new ("[kernel]" ,0 ); -if (!kernel_dso) -return -1 ; - - err =dso__load_kernel (kernel_dso, vmlinux,NULL , verbose, modules); -if (err <=0 ) { -dso__delete (kernel_dso); - kernel_dso =NULL ; - }else -dsos__add (kernel_dso); - - vdso =dso__new ("[vdso]" ,0 ); -if (!vdso) -return -1 ; - - vdso->find_symbol = vdso__find_symbol; - -dsos__add (vdso); - -return err; -} -struct map {struct list_head node;u64 start;tools/perf/builtin-record.c +0
-1@@ -40,7 +40,6 @@ static int inherit = 1;static int force =0 ;static int append_file =0 ;static int call_graph =0 ; -static int verbose =0 ;static int inherit_stat =0 ;static int no_samples =0 ;static int sample_address =0 ;tools/perf/builtin-report.c +0
-97@@ -30,7 +30,6 @@#define SHOW_HV 4static char const *input_name ="perf.data" ; -static char *vmlinux =NULL ;static char default_sort_order[] ="comm,dso,symbol" ;static char *sort_order = default_sort_order;@@ -46,11 +45,6 @@ static int dump_trace = 0; #define dprintf(x...) do { if (dump_trace) printf(x); } while (0) #define cdprintf(x...) do { if (dump_trace) color_fprintf(stdout, color, x); } while (0) -static int verbose; -#define eprintf(x...) do { if (verbose) fprintf(stderr, x); } while (0) - -static int modules; -static int full_paths;static int show_nr_samples;@@ -161,98 +155,7 @@ static int repsep_fprintf(FILE *fp, const char *fmt, ...)return n; } -static LIST_HEAD (dsos); -static struct dso *kernel_dso; -static struct dso *vdso; -static struct dso *hypervisor_dso; - -static void dsos__add (struct dso *dso) -{ -list_add_tail (&dso->node, &dsos); -} - -static struct dso *dsos__find (const char *name) -{ -struct dso *pos; - -list_for_each_entry (pos, &dsos, node) -if (strcmp (pos->name, name) ==0 ) -return pos; -return NULL ; -} - -static struct dso *dsos__findnew (const char *name) -{ -struct dso *dso =dsos__find (name); -int nr; - -if (dso) -return dso; - - dso =dso__new (name,0 ); -if (!dso) - goto out_delete_dso; - - nr =dso__load (dso,NULL , verbose); -if (nr <0 ) { -eprintf ("Failed to open: %s\n" , name); - goto out_delete_dso; - } -if (!nr) -eprintf ("No symbols found in: %s, maybe install a debug package?\n" , name); - -dsos__add (dso); - -return dso; - -out_delete_dso: -dso__delete (dso); -return NULL ; -} - -static void dsos__fprintf (FILE *fp) -{ -struct dso *pos; - -list_for_each_entry (pos, &dsos, node) -dso__fprintf (pos, fp); -} - -static struct symbol *vdso__find_symbol (struct dso *dso,u64 ip) -{ -return dso__find_symbol (dso, ip); -} - -static int load_kernel (void ) -{ -int err; - - kernel_dso =dso__new ("[kernel]" ,0 ); -if (!kernel_dso) -return -1 ; - - err =dso__load_kernel (kernel_dso, vmlinux,NULL , verbose, modules); -if (err <=0 ) { -dso__delete (kernel_dso); - kernel_dso =NULL ; - }else -dsos__add (kernel_dso); - - vdso =dso__new ("[vdso]" ,0 ); -if (!vdso) -return -1 ; - vdso->find_symbol = vdso__find_symbol; - -dsos__add (vdso); - - hypervisor_dso =dso__new ("[hypervisor]" ,0 ); -if (!hypervisor_dso) -return -1 ; -dsos__add (hypervisor_dso); - -return err; -}static char __cwd[PATH_MAX ];static char *cwd = __cwd;tools/perf/builtin-stat.c +0
-1@@ -63,7 +63,6 @@ static struct perf_counter_attr default_attrs[] = {#define MAX_RUN 100static int system_wide =0 ; -static int verbose =0 ;static unsigned int nr_cpus =0 ;static int run_idx =0 ;tools/perf/builtin-top.c +0
-4@@ -68,8 +68,6 @@ static int group = 0;static unsigned int page_size;static unsigned int mmap_pages =16 ;static int freq =0 ; -static int verbose =0 ; -static char *vmlinux =NULL ;static int delay_secs =2 ;static int zero;@@ -338,8 +336,6 @@ static void show_details(struct sym_entry *syme)printf ("%d lines not displayed, maybe increase display entries [e]\n" , more); } -struct dso *kernel_dso; -/* * Symbols will be added here in record_ip and will get out * after decayed. tools/perf/builtin.h +1
-0@@ -22,5 +22,6 @@ extern int cmd_stat(int argc, const char **argv, const char *prefix);extern int cmd_top (int argc,const char **argv,const char *prefix);extern int cmd_version (int argc,const char **argv,const char *prefix);extern int cmd_list (int argc,const char **argv,const char *prefix); +extern int cmd_trace (int argc,const char **argv,const char *prefix);#endif tools/perf/perf.h +1
-0@@ -48,6 +48,7 @@#include "../../include/linux/perf_counter.h" #include "util/types.h" +#include "util/debug.h" /* * prctl(PR_TASK_PERF_COUNTERS_DISABLE) will (cheaply) disable all tools/perf/util/symbol.c +97
-0@@ -937,6 +937,103 @@ int dso__load_kernel(struct dso *self, const char *vmlinux,return err; } +LIST_HEAD (dsos); +struct dso *kernel_dso; +struct dso *vdso; +struct dso *hypervisor_dso; + +char *vmlinux ="vmlinux" ; +int modules; + +static void dsos__add (struct dso *dso) +{ +list_add_tail (&dso->node, &dsos); +} + +static struct dso *dsos__find (const char *name) +{ +struct dso *pos; + +list_for_each_entry (pos, &dsos, node) +if (strcmp (pos->name, name) ==0 ) +return pos; +return NULL ; +} + +struct dso *dsos__findnew (const char *name) +{ +struct dso *dso =dsos__find (name); +int nr; + +if (dso) +return dso; + + dso =dso__new (name,0 ); +if (!dso) + goto out_delete_dso; + + nr =dso__load (dso,NULL , verbose); +if (nr <0 ) { +eprintf ("Failed to open: %s\n" , name); + goto out_delete_dso; + } +if (!nr) +eprintf ("No symbols found in: %s, maybe install a debug package?\n" , name); + +dsos__add (dso); + +return dso; + +out_delete_dso: +dso__delete (dso); +return NULL ; +} + +void dsos__fprintf (FILE *fp) +{ +struct dso *pos; + +list_for_each_entry (pos, &dsos, node) +dso__fprintf (pos, fp); +} + +static struct symbol *vdso__find_symbol (struct dso *dso,u64 ip) +{ +return dso__find_symbol (dso, ip); +} + +int load_kernel (void ) +{ +int err; + + kernel_dso =dso__new ("[kernel]" ,0 ); +if (!kernel_dso) +return -1 ; + + err =dso__load_kernel (kernel_dso, vmlinux,NULL , verbose, modules); +if (err <=0 ) { +dso__delete (kernel_dso); + kernel_dso =NULL ; + }else +dsos__add (kernel_dso); + + vdso =dso__new ("[vdso]" ,0 ); +if (!vdso) +return -1 ; + + vdso->find_symbol = vdso__find_symbol; + +dsos__add (vdso); + + hypervisor_dso =dso__new ("[hypervisor]" ,0 ); +if (!hypervisor_dso) +return -1 ; +dsos__add (hypervisor_dso); + +return err; +} + +void symbol__init (void ) {elf_version (EV_CURRENT );tools/perf/util/symbol.h +11
-0@@ -48,9 +48,20 @@ int dso__load_kernel(struct dso *self, const char *vmlinux, symbol_filter_t filter, int verbose, int modules);int dso__load_modules (struct dso *self,symbol_filter_t filter,int verbose);int dso__load (struct dso *self,symbol_filter_t filter,int verbose); +struct dso *dsos__findnew (const char *name); +void dsos__fprintf (FILE *fp);size_t dso__fprintf (struct dso *self,FILE *fp);char dso__symtab_origin (const struct dso *self); +int load_kernel (void ); +void symbol__init (void ); + +extern struct list_head dsos; +extern struct dso *kernel_dso; +extern struct dso *vdso; +extern struct dso *hypervisor_dso; +extern char *vmlinux; +extern int modules;#endif /* _PERF_SYMBOL_ */ Signature · 1 line
1.6.2.3
Re: [PATCH 1/3] perf tools: Factorize high level dso helpers
Quoted 38 lines · Frederic Weisbecker
Quoted 29 lines · Frederic Weisbecker
Factorize multiple definitions of high level dso helpers into the symbol source file.
The side effect is a general export of the verbose and eprintf debugging helpers into a new file dedicated to debugging purposes.
Signed-off-by Frederic Weisbecker <[email protected]>Cc Arnaldo Carvalho de Melo <[email protected]>Cc Peter Zijlstra <[email protected]>Cc Mike Galbraith <[email protected]>Cc Brice Goglin <[email protected]>--- tools/perf/Makefile | 1 + tools/perf/builtin-annotate.c | 96 ---------------------------------------- tools/perf/builtin-record.c | 1 - tools/perf/builtin-report.c | 97 ----------------------------------------- tools/perf/builtin-stat.c | 1 - tools/perf/builtin-top.c | 4 -- tools/perf/builtin.h | 1 + tools/perf/perf.h | 1 + tools/perf/util/symbol.c | 97 +++++++++++++++++++++++++++++++++++++++++ tools/perf/util/symbol.h | 11 +++++ 10 files changed, 111 insertions(+), 199 deletions(-)tools/perf/Makefile +0
-0tools/perf/Makefile LIB_OBJS += util/pager.o · line 342
Quoted 4 lines
tools/perf/Makefile +1
-0@@ -339,6 +339,7 @@ LIB_OBJS += util/pager.o LIB_OBJS += util/header.o LIB_OBJS += util/callchain.o LIB_OBJS += util/values.o +LIB_OBJS += util/debug.oDamn, did I forgot to git-add
util/debug.c?
Quoted 1 line
How could I fix this? Another pull request?
yeah. I've zapped those commits from perfcounters/core.
Ingo
Quoted 3 lines · Ingo Molnar
yeah. I've zapped those commits from perfcounters/core.
Ingo
Ok, sorry about that.
I've rebuilt a clean tree. It should be fine now.
Thanks.
The following changes since commit 9f8666971185b86615a074bcac67c90fdf8af8bc:
Brice Goglin (1):
perf report: Add raw displaying of per-thread counters
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing.git \
perfcounters/core
Frederic Weisbecker (3):
perf tools: Factorize high level dso helpers
perf tools: Factorize the event structure definitions in a single file perf tools: Factorize the map helpers
tools/perf/Makefile | 2 + tools/perf/builtin-annotate.c | 209 +------------------------------- tools/perf/builtin-record.c | 20 +--- tools/perf/builtin-report.c | 274 +---------------------------------------- tools/perf/builtin-stat.c | 1 - tools/perf/builtin-top.c | 24 ---- tools/perf/builtin.h | 1 + tools/perf/perf.h | 1 + tools/perf/util/callchain.h | 1 + tools/perf/util/debug.c | 22 ++++ tools/perf/util/debug.h | 5 + tools/perf/util/event.h | 84 +++++++++++++ tools/perf/util/map.c | 97 +++++++++++++++ tools/perf/util/symbol.c | 97 +++++++++++++++ tools/perf/util/symbol.h | 12 ++ tools/perf/util/util.h | 1 + 16 files changed, 330 insertions(+), 521 deletions(-) create mode 100644 tools/perf/util/debug.c create mode 100644 tools/perf/util/debug.h create mode 100644 tools/perf/util/event.h create mode 100644 tools/perf/util/map.c
Quoted 7 lines · Frederic Weisbecker
Quoted 3 lines · Ingo Molnar
yeah. I've zapped those commits from perfcounters/core.
Ingo
Ok, sorry about that.
I've rebuilt a clean tree. It should be fine now.
btw., you can always check whether there's any missing git-add, via:
git ls-files --others
And the safest way to commit is to use 'git commit -a' - that will auto-include new files. (but the tree has to be clean for that)
Quoted 30 lines
Thanks.
The following changes since commit
9f8666971185b86615a074bcac67c90fdf8af8bc:
Brice Goglin (1):
perf report: Add raw displaying of per-thread countersare available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing.git\
perfcounters/coreFrederic Weisbecker (3):
perf tools: Factorize high level dso helpers
perf tools: Factorize the event structure definitions in a single file perf tools: Factorize the map helperstools/perf/Makefile | 2 + tools/perf/builtin-annotate.c | 209 +------------------------------- tools/perf/builtin-record.c | 20 +--- tools/perf/builtin-report.c | 274 +---------------------------------------- tools/perf/builtin-stat.c | 1 - tools/perf/builtin-top.c | 24 ---- tools/perf/builtin.h | 1 + tools/perf/perf.h | 1 + tools/perf/util/callchain.h | 1 + tools/perf/util/debug.c | 22 ++++ tools/perf/util/debug.h | 5 + tools/perf/util/event.h | 84 +++++++++++++ tools/perf/util/map.c | 97 +++++++++++++++
Quoted 8 lines
tools/perf/util/symbol.c | 97 +++++++++++++++ tools/perf/util/symbol.h | 12 ++ tools/perf/util/util.h | 1 + 16 files changed, 330 insertions(+), 521 deletions(-) create mode 100644 tools/perf/util/debug.c create mode 100644 tools/perf/util/debug.h create mode 100644 tools/perf/util/event.h create mode 100644 tools/perf/util/map.c
Re-pulled it - it works fine here too - thanks!
Ingo
Quoted 8 lines · Ingo Molnar
Quoted 5 lines · Frederic Weisbecker
Quoted 3 lines · Ingo Molnar
yeah. I've zapped those commits from perfcounters/core.
Ingo
Quoted 7 lines
Quoted 2 lines
Ok, sorry about that.
I've rebuilt a clean tree. It should be fine now.btw., you can always check whether there's any missing git-add, via:
git ls-files --others
Ah nice tip.
Quoted 3 lines
And the safest way to commit is to use 'git commit -a' - that will auto-include new files. (but the tree has to be clean for that)
Well, I always have junks in my directories. .rej/.orig files, or more especially scripts and tests in the case of the tools/perf directory. So it's hard to check everything this way.
Note that most of the time I use git-commit -a, but it doesn't include untracked files AFAICS.
Quoted 39 lines
Quoted 30 lines
Thanks.
The following changes since commit
9f8666971185b86615a074bcac67c90fdf8af8bc:
Brice Goglin (1):
perf report: Add raw displaying of per-thread countersare available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing.git\
perfcounters/coreFrederic Weisbecker (3):
perf tools: Factorize high level dso helpers
perf tools: Factorize the event structure definitions in a single file perf tools: Factorize the map helperstools/perf/Makefile | 2 + tools/perf/builtin-annotate.c | 209 +------------------------------- tools/perf/builtin-record.c | 20 +--- tools/perf/builtin-report.c | 274 +---------------------------------------- tools/perf/builtin-stat.c | 1 - tools/perf/builtin-top.c | 24 ---- tools/perf/builtin.h | 1 + tools/perf/perf.h | 1 + tools/perf/util/callchain.h | 1 + tools/perf/util/debug.c | 22 ++++ tools/perf/util/debug.h | 5 + tools/perf/util/event.h | 84 +++++++++++++ tools/perf/util/map.c | 97 +++++++++++++++Quoted 8 lines
tools/perf/util/symbol.c | 97 +++++++++++++++ tools/perf/util/symbol.h | 12 ++ tools/perf/util/util.h | 1 + 16 files changed, 330 insertions(+), 521 deletions(-) create mode 100644 tools/perf/util/debug.c create mode 100644 tools/perf/util/debug.h create mode 100644 tools/perf/util/event.h create mode 100644 tools/perf/util/map.c
Quoted 3 lines
Re-pulled it - it works fine here too - thanks!
Ingo
Cool, thanks.
Em Wed, Aug 12, 2009 at 12:48:29PM +0200, Frederic Weisbecker escreveu:
Quoted 9 lines · Ingo Molnar
Quoted 3 lines
yeah. I've zapped those commits from perfcounters/core.
Ingo
Ok, sorry about that.
I've rebuilt a clean tree. It should be fine now.Thanks.
Thanks a lot for doing this work, I intended to do that but got sidetracked with symbol resolving neverending corner cases 8)
- Arnaldo
Quoted 6 lines · Arnaldo Carvalho de Melo
Em Wed, Aug 12, 2009 at 12:48:29PM +0200, Frederic Weisbecker escreveu:
Quoted 5 lines · Ingo Molnar
Quoted 3 lines
yeah. I've zapped those commits from perfcounters/core.
Ingo
Quoted 7 lines
Quoted 4 lines
Ok, sorry about that.
I've rebuilt a clean tree. It should be fine now.Thanks.
Thanks a lot for doing this work, I intended to do that but got sidetracked with symbol resolving neverending corner cases 8)
:)
Also, there's still quite a bit of code left to be factorized out of builtin-annotate.c and builtin-report.c.
Ingo
PATCH v2 1/3 perf tools: Factorize high level dso helpers
In reply to Ingo MolnarFactorize multiple definitions of high level dso helpers into the symbol source file.
The side effect is a general export of the verbose and eprintf debugging helpers into a new file dedicated to debugging purposes.
--- tools/perf/Makefile | 1 + tools/perf/builtin-annotate.c | 96 ---------------------------------------- tools/perf/builtin-record.c | 1 - tools/perf/builtin-report.c | 97 ----------------------------------------- tools/perf/builtin-stat.c | 1 - tools/perf/builtin-top.c | 4 -- tools/perf/builtin.h | 1 + tools/perf/perf.h | 1 + tools/perf/util/debug.c | 22 +++++++++ tools/perf/util/debug.h | 5 ++ tools/perf/util/symbol.c | 97 +++++++++++++++++++++++++++++++++++++++++ tools/perf/util/symbol.h | 11 +++++ 12 files changed, 138 insertions(+), 199 deletions(-) create mode 100644 tools/perf/util/debug.c create mode 100644 tools/perf/util/debug.h
tools/perf/Makefile +1 -0
@@ -339,6 +339,7 @@ LIB_OBJS += util/pager.o LIB_OBJS += util/header.o LIB_OBJS += util/callchain.o LIB_OBJS += util/values.o +LIB_OBJS += util/debug.o BUILTIN_OBJS += builtin-annotate.o BUILTIN_OBJS += builtin-help.o
tools/perf/builtin-annotate.c +0 -96
114 lines not on this page · raw message
tools/perf/builtin-record.c +0 -1
@@ -40,7 +40,6 @@ static int inherit = 1;static int force =0 ;static int append_file =0 ;static int call_graph =0 ; -static int verbose =0 ;static int inherit_stat =0 ;static int no_samples =0 ;static int sample_address =0 ;
tools/perf/builtin-report.c +0 -97
116 lines not on this page · raw message
tools/perf/builtin-stat.c +0 -1
@@ -63,7 +63,6 @@ static struct perf_counter_attr default_attrs[] = {#define MAX_RUN 100static int system_wide =0 ; -static int verbose =0 ;static unsigned int nr_cpus =0 ;static int run_idx =0 ;
tools/perf/builtin-top.c +0 -4
@@ -68,8 +68,6 @@ static int group = 0;static unsigned int page_size;static unsigned int mmap_pages =16 ;static int freq =0 ; -static int verbose =0 ; -static char *vmlinux =NULL ;static int delay_secs =2 ;static int zero;
@@ -338,8 +336,6 @@ static void show_details(struct sym_entry *syme)printf ("%d lines not displayed, maybe increase display entries [e]\n" , more); } -struct dso *kernel_dso; -/* * Symbols will be added here in record_ip and will get out * after decayed.
tools/perf/builtin.h +1 -0
@@ -22,5 +22,6 @@ extern int cmd_stat(int argc, const char **argv, const char *prefix);extern int cmd_top (int argc,const char **argv,const char *prefix);extern int cmd_version (int argc,const char **argv,const char *prefix);extern int cmd_list (int argc,const char **argv,const char *prefix); +extern int cmd_trace (int argc,const char **argv,const char *prefix);#endif
tools/perf/perf.h +1 -0
@@ -48,6 +48,7 @@#include "../../include/linux/perf_counter.h" #include "util/types.h" +#include "util/debug.h" /* * prctl(PR_TASK_PERF_COUNTERS_DISABLE) will (cheaply) disable all
tools/perf/util/debug.c new file +22 -0
@@ -0,0 +1,22 @@ +/* For general debugging purposes */ + +#include "../perf.h" +#include <string.h> +#include <stdarg.h> +#include <stdio.h> + +int verbose =0 ; + +int eprintf (const char *fmt, ...) +{ +va_list args; +int ret =0 ; + +if (verbose) { +va_start (args, fmt); + ret =vfprintf (stderr, fmt, args); +va_end (args); + } + +return ret; +}
tools/perf/util/debug.h new file +5 -0
@@ -0,0 +1,5 @@ +/* For debugging general purposes */ + +extern int verbose; + +int eprintf (const char *fmt, ...) __attribute__((format (printf,1 ,2 )));
tools/perf/util/symbol.c +97 -0
103 lines not on this page · raw message
tools/perf/util/symbol.h +11 -0
@@ -48,9 +48,20 @@ int dso__load_kernel(struct dso *self, const char *vmlinux, symbol_filter_t filter, int verbose, int modules);int dso__load_modules (struct dso *self,symbol_filter_t filter,int verbose);int dso__load (struct dso *self,symbol_filter_t filter,int verbose); +struct dso *dsos__findnew (const char *name); +void dsos__fprintf (FILE *fp);size_t dso__fprintf (struct dso *self,FILE *fp);char dso__symtab_origin (const struct dso *self); +int load_kernel (void ); +void symbol__init (void ); + +extern struct list_head dsos; +extern struct dso *kernel_dso; +extern struct dso *vdso; +extern struct dso *hypervisor_dso; +extern char *vmlinux; +extern int modules;#endif /* _PERF_SYMBOL_ */
Signature · 1 line
1.6.2.3
PATCH v2 2/3 perf tools: Factorize the event structure definitions in a single file
In reply to Ingo MolnarFactorize the multiple definition of the events structures into a single util/event.h file.
--- tools/perf/builtin-annotate.c | 34 ------------------------- tools/perf/builtin-record.c | 18 ------------- tools/perf/builtin-report.c | 53 ---------------------------------------- tools/perf/builtin-top.c | 20 --------------- tools/perf/util/event.h | 54 +++++++++++++++++++++++++++++++++++++++++ tools/perf/util/util.h | 2 + 6 files changed, 56 insertions(+), 125 deletions(-) create mode 100644 tools/perf/util/event.h
tools/perf/builtin-annotate.c +0 -34
@@ -44,40 +44,6 @@ static int print_line;static unsigned long page_size;static unsigned long mmap_window =32 ; -struct ip_event { -struct perf_event_header header; -u64 ip; -u32 pid, tid; -}; - -struct mmap_event { -struct perf_event_header header; -u32 pid, tid; -u64 start; -u64 len; -u64 pgoff; -char filename[PATH_MAX ]; -}; - -struct comm_event { -struct perf_event_header header; -u32 pid, tid; -char comm[16 ]; -}; - -struct fork_event { -struct perf_event_header header; -u32 pid, ppid; -}; - -typedef union event_union { -struct perf_event_header header; -struct ip_event ip; -struct mmap_event mmap; -struct comm_event comm; -struct fork_event fork; -}event_t ; -struct sym_ext {struct rb_node node;
tools/perf/builtin-record.c +0 -18
@@ -59,24 +59,6 @@ static int file_new = 1;struct perf_header *header; -struct mmap_event { -struct perf_event_header header; -u32 pid; -u32 tid; -u64 start; -u64 len; -u64 pgoff; -char filename[PATH_MAX ]; -}; - -struct comm_event { -struct perf_event_header header; -u32 pid; -u32 tid; -char comm[16 ]; -}; - -struct mmap_data {int counter;void *base;
tools/perf/builtin-report.c +0 -53
59 lines not on this page · raw message
tools/perf/builtin-top.c +0 -20
@@ -933,26 +933,6 @@ static void mmap_read_counter(struct mmap_data *md) last_read = this_read;for (; old != head;) { -struct ip_event { -struct perf_event_header header; -u64 ip; -u32 pid, target_pid; - }; -struct mmap_event { -struct perf_event_header header; -u32 pid, target_pid; -u64 start; -u64 len; -u64 pgoff; -char filename[PATH_MAX ]; - }; - -typedef union event_union { -struct perf_event_header header; -struct ip_event ip; -struct mmap_event mmap; - }event_t ; -event_t *event = (event_t *)&data[old & md->mask];event_t event_copy;
tools/perf/util/event.h new file +54 -0
54 lines not on this page · raw message
tools/perf/util/util.h +2 -0
@@ -83,6 +83,8 @@#include <inttypes.h> #include "../../../include/linux/magic.h" +#include "event.h" +#ifndef NO_ICONV #include <iconv.h> #endif
Signature · 1 line
1.6.2.3
PATCH v2 3/3 perf tools: Factorize the map helpers
In reply to Ingo MolnarFactorize the dso mapping helpers into a single purpose common file "util/map.c"
--- tools/perf/Makefile | 1 + tools/perf/builtin-annotate.c | 79 +-------------------------- tools/perf/builtin-record.c | 1 + tools/perf/builtin-report.c | 124 ++--------------------------------------- tools/perf/util/callchain.h | 1 + tools/perf/util/event.h | 30 ++++++++++ tools/perf/util/map.c | 97 ++++++++++++++++++++++++++++++++ tools/perf/util/symbol.h | 1 + tools/perf/util/util.h | 1 - 9 files changed, 137 insertions(+), 198 deletions(-) create mode 100644 tools/perf/util/map.c
tools/perf/Makefile +1 -0
@@ -340,6 +340,7 @@ LIB_OBJS += util/header.o LIB_OBJS += util/callchain.o LIB_OBJS += util/values.o LIB_OBJS += util/debug.o +LIB_OBJS += util/map.o BUILTIN_OBJS += builtin-annotate.o BUILTIN_OBJS += builtin-help.o
tools/perf/builtin-annotate.c +1 -78
91 lines not on this page · raw message
tools/perf/builtin-record.c +1 -0
@@ -15,6 +15,7 @@#include "util/string.h" #include "util/header.h" +#include "util/event.h" #include <unistd.h> #include <sched.h>
tools/perf/builtin-report.c +5 -119
142 lines not on this page · raw message
tools/perf/util/callchain.h +1 -0
@@ -4,6 +4,7 @@#include "../perf.h" #include <linux/list.h> #include <linux/rbtree.h> +#include "util.h" #include "symbol.h" enum chain_mode {
tools/perf/util/event.h +30 -0
@@ -1,4 +1,8 @@ +#ifndef __PERF_EVENT_H +#define __PERF_EVENT_H #include "../perf.h" +#include "util.h" +#include <linux/list.h> struct ip_event { struct perf_event_header header;
@@ -52,3 +56,29 @@ typedef union event_union {struct lost_event lost;struct read_event read; } event_t; + +struct map { +struct list_head node; +u64 start; +u64 end; +u64 pgoff; +u64 (*map_ip)(struct map *,u64 ); +struct dso *dso; +}; + +static inline u64 map__map_ip (struct map *map,u64 ip) +{ +return ip - map->start + map->pgoff; +} + +static inline u64 vdso__map_ip (struct map *map__used ,u64 ip) +{ +return ip; +} + +struct map *map__new (struct mmap_event *event,char *cwd,int cwdlen); +struct map *map__clone (struct map *self); +int map__overlap (struct map *l,struct map *r); +size_t map__fprintf (struct map *self,FILE *fp); + +#endif
tools/perf/util/map.c new file +97 -0
97 lines not on this page · raw message
tools/perf/util/symbol.h +1 -0
@@ -6,6 +6,7 @@#include <linux/list.h> #include <linux/rbtree.h> #include "module.h" +#include "event.h" struct symbol {struct rb_node rb_node;
tools/perf/util/util.h +0 -1
@@ -83,7 +83,6 @@#include <inttypes.h> #include "../../../include/linux/magic.h" -#include "event.h" #ifndef NO_ICONV #include <iconv.h>
Signature · 1 line
1.6.2.3
PATCH 2/3 perf tools: Factorize the event structure definitions in a single file
In reply to Frederic WeisbeckerFactorize the multiple definition of the events structures into a single util/event.h file.
--- tools/perf/builtin-annotate.c | 34 ------------------------- tools/perf/builtin-record.c | 18 ------------- tools/perf/builtin-report.c | 53 ---------------------------------------- tools/perf/builtin-top.c | 20 --------------- tools/perf/util/event.h | 54 +++++++++++++++++++++++++++++++++++++++++ tools/perf/util/util.h | 2 + 6 files changed, 56 insertions(+), 125 deletions(-) create mode 100644 tools/perf/util/event.h
tools/perf/builtin-annotate.c +0 -34
@@ -44,40 +44,6 @@ static int print_line;static unsigned long page_size;static unsigned long mmap_window =32 ; -struct ip_event { -struct perf_event_header header; -u64 ip; -u32 pid, tid; -}; - -struct mmap_event { -struct perf_event_header header; -u32 pid, tid; -u64 start; -u64 len; -u64 pgoff; -char filename[PATH_MAX ]; -}; - -struct comm_event { -struct perf_event_header header; -u32 pid, tid; -char comm[16 ]; -}; - -struct fork_event { -struct perf_event_header header; -u32 pid, ppid; -}; - -typedef union event_union { -struct perf_event_header header; -struct ip_event ip; -struct mmap_event mmap; -struct comm_event comm; -struct fork_event fork; -}event_t ; -struct sym_ext {struct rb_node node;
tools/perf/builtin-record.c +0 -18
@@ -59,24 +59,6 @@ static int file_new = 1;struct perf_header *header; -struct mmap_event { -struct perf_event_header header; -u32 pid; -u32 tid; -u64 start; -u64 len; -u64 pgoff; -char filename[PATH_MAX ]; -}; - -struct comm_event { -struct perf_event_header header; -u32 pid; -u32 tid; -char comm[16 ]; -}; - -struct mmap_data {int counter;void *base;
tools/perf/builtin-report.c +0 -53
59 lines not on this page · raw message
tools/perf/builtin-top.c +0 -20
@@ -933,26 +933,6 @@ static void mmap_read_counter(struct mmap_data *md) last_read = this_read;for (; old != head;) { -struct ip_event { -struct perf_event_header header; -u64 ip; -u32 pid, target_pid; - }; -struct mmap_event { -struct perf_event_header header; -u32 pid, target_pid; -u64 start; -u64 len; -u64 pgoff; -char filename[PATH_MAX ]; - }; - -typedef union event_union { -struct perf_event_header header; -struct ip_event ip; -struct mmap_event mmap; - }event_t ; -event_t *event = (event_t *)&data[old & md->mask];event_t event_copy;
tools/perf/util/event.h new file +54 -0
54 lines not on this page · raw message
tools/perf/util/util.h +2 -0
@@ -83,6 +83,8 @@#include <inttypes.h> #include "../../../include/linux/magic.h" +#include "event.h" +#ifndef NO_ICONV #include <iconv.h> #endif
Signature · 1 line
1.6.2.3
PATCH 3/3 perf tools: Factorize the map helpers
In reply to Frederic WeisbeckerFactorize the dso mapping helpers into a single purpose common file "util/map.c"
--- tools/perf/Makefile | 1 + tools/perf/builtin-annotate.c | 79 +-------------------------- tools/perf/builtin-record.c | 1 + tools/perf/builtin-report.c | 124 ++--------------------------------------- tools/perf/util/callchain.h | 1 + tools/perf/util/event.h | 30 ++++++++++ tools/perf/util/map.c | 97 ++++++++++++++++++++++++++++++++ tools/perf/util/symbol.h | 1 + tools/perf/util/util.h | 1 - 9 files changed, 137 insertions(+), 198 deletions(-) create mode 100644 tools/perf/util/map.c
tools/perf/Makefile +1 -0
@@ -340,6 +340,7 @@ LIB_OBJS += util/header.o LIB_OBJS += util/callchain.o LIB_OBJS += util/values.o LIB_OBJS += util/debug.o +LIB_OBJS += util/event.o BUILTIN_OBJS += builtin-annotate.o BUILTIN_OBJS += builtin-help.o
tools/perf/builtin-annotate.c +1 -78
91 lines not on this page · raw message
tools/perf/builtin-record.c +1 -0
@@ -15,6 +15,7 @@#include "util/string.h" #include "util/header.h" +#include "util/event.h" #include <unistd.h> #include <sched.h>
tools/perf/builtin-report.c +5 -119
142 lines not on this page · raw message
tools/perf/util/callchain.h +1 -0
@@ -4,6 +4,7 @@#include "../perf.h" #include <linux/list.h> #include <linux/rbtree.h> +#include "util.h" #include "symbol.h" enum chain_mode {
tools/perf/util/event.h +30 -0
@@ -1,4 +1,8 @@ +#ifndef __PERF_EVENT_H +#define __PERF_EVENT_H #include "../perf.h" +#include "util.h" +#include <linux/list.h> struct ip_event { struct perf_event_header header;
@@ -52,3 +56,29 @@ typedef union event_union {struct lost_event lost;struct read_event read; } event_t; + +struct map { +struct list_head node; +u64 start; +u64 end; +u64 pgoff; +u64 (*map_ip)(struct map *,u64 ); +struct dso *dso; +}; + +static inline u64 map__map_ip (struct map *map,u64 ip) +{ +return ip - map->start + map->pgoff; +} + +static inline u64 vdso__map_ip (struct map *map__used ,u64 ip) +{ +return ip; +} + +struct map *map__new (struct mmap_event *event,char *cwd,int cwdlen); +struct map *map__clone (struct map *self); +int map__overlap (struct map *l,struct map *r); +size_t map__fprintf (struct map *self,FILE *fp); + +#endif
tools/perf/util/map.c new file +97 -0
97 lines not on this page · raw message
tools/perf/util/symbol.h +1 -0
@@ -6,6 +6,7 @@#include <linux/list.h> #include <linux/rbtree.h> #include "module.h" +#include "event.h" struct symbol {struct rb_node rb_node;
tools/perf/util/util.h +0 -1
@@ -83,7 +83,6 @@#include <inttypes.h> #include "../../../include/linux/magic.h" -#include "event.h" #ifndef NO_ICONV #include <iconv.h>
Signature · 1 line
1.6.2.3
Quoted 17 lines · Frederic Weisbecker
Ingo,
This set factorizes the common definition of structs, functions, variables inside perf tools to make it easier for perf trace to be created.
I guess this patchset only factorizes 10 % of what could be done, the largest part is the intersection between perf report and perf annotate. It's a lot of work.
I guess I will do the rest in separate patchsets while writing perf trace.
Thanks,
Frederic.
Quoted 7 lines
The following changes since commit
9f8666971185b86615a074bcac67c90fdf8af8bc:
Brice Goglin (1):
perf report: Add raw displaying of per-thread countersare available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing.git..BRANCH.NOT.VERIFIED..
s/BRANCH.NOT.VERIFIED/"perfcounters/core"
Quoted 23 lines
Frederic Weisbecker (3):
perf tools: Factorize high level dso helpers
perf tools: Factorize the event structure definitions in a single file perf tools: Factorize the map helperstools/perf/Makefile | 2 + tools/perf/builtin-annotate.c | 209 +------------------------------- tools/perf/builtin-record.c | 20 +--- tools/perf/builtin-report.c | 274 +---------------------------------------- tools/perf/builtin-stat.c | 1 - tools/perf/builtin-top.c | 24 ---- tools/perf/builtin.h | 1 + tools/perf/perf.h | 1 + tools/perf/util/callchain.h | 1 + tools/perf/util/event.h | 84 +++++++++++++ tools/perf/util/map.c | 97 +++++++++++++++ tools/perf/util/symbol.c | 97 +++++++++++++++ tools/perf/util/symbol.h | 12 ++ tools/perf/util/util.h | 1 + 14 files changed, 303 insertions(+), 521 deletions(-) create mode 100644 tools/perf/util/event.h create mode 100644 tools/perf/util/map.c
Quoted 6 lines · Frederic Weisbecker
Ingo,
This set factorizes the common definition of structs, functions, variables inside perf tools to make it easier for perf trace to be created.
Quoted 6 lines
I guess this patchset only factorizes 10 % of what could be done, the largest part is the intersection between perf report and perf annotate. It's a lot of work.
I guess I will do the rest in separate patchsets while writing perf trace.
Pulled, thanks Frederic!
Ingo
Peter Zijlstra
maintainerQuoted 6 lines · Frederic Weisbecker
Ingo,
This set factorizes the common definition of structs, functions, variables inside perf tools to make it easier for perf trace to be created.
Quoted 6 lines
I guess this patchset only factorizes 10 % of what could be done, the largest part is the intersection between perf report and perf annotate. It's a lot of work.
I guess I will do the rest in separate patchsets while writing perf trace.
Awesome, much needed cleanups.
Thanks