2009-03-22 09:31:04 +01:00
|
|
|
/*
|
|
|
|
* trace_events_filter - generic event filtering
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or modify
|
|
|
|
* it under the terms of the GNU General Public License as published by
|
|
|
|
* the Free Software Foundation; either version 2 of the License, or
|
|
|
|
* (at your option) any later version.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope that it will be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
|
|
* GNU General Public License for more details.
|
|
|
|
*
|
|
|
|
* You should have received a copy of the GNU General Public License
|
|
|
|
* along with this program; if not, write to the Free Software
|
|
|
|
* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
|
|
|
|
*
|
|
|
|
* Copyright (C) 2009 Tom Zanussi <tzanussi@gmail.com>
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/ctype.h>
|
2009-04-17 07:27:08 +02:00
|
|
|
#include <linux/mutex.h>
|
2009-10-15 05:21:42 +02:00
|
|
|
#include <linux/perf_event.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 09:04:11 +01:00
|
|
|
#include <linux/slab.h>
|
2009-03-22 09:31:04 +01:00
|
|
|
|
|
|
|
#include "trace.h"
|
2009-03-24 08:14:31 +01:00
|
|
|
#include "trace_output.h"
|
2009-03-22 09:31:04 +01:00
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
enum filter_op_ids
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
OP_OR,
|
|
|
|
OP_AND,
|
2009-10-15 05:21:12 +02:00
|
|
|
OP_GLOB,
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
OP_NE,
|
|
|
|
OP_EQ,
|
|
|
|
OP_LT,
|
|
|
|
OP_LE,
|
|
|
|
OP_GT,
|
|
|
|
OP_GE,
|
|
|
|
OP_NONE,
|
|
|
|
OP_OPEN_PAREN,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct filter_op {
|
|
|
|
int id;
|
|
|
|
char *string;
|
|
|
|
int precedence;
|
|
|
|
};
|
|
|
|
|
|
|
|
static struct filter_op filter_ops[] = {
|
2009-10-15 05:21:12 +02:00
|
|
|
{ OP_OR, "||", 1 },
|
|
|
|
{ OP_AND, "&&", 2 },
|
|
|
|
{ OP_GLOB, "~", 4 },
|
|
|
|
{ OP_NE, "!=", 4 },
|
|
|
|
{ OP_EQ, "==", 4 },
|
|
|
|
{ OP_LT, "<", 5 },
|
|
|
|
{ OP_LE, "<=", 5 },
|
|
|
|
{ OP_GT, ">", 5 },
|
|
|
|
{ OP_GE, ">=", 5 },
|
|
|
|
{ OP_NONE, "OP_NONE", 0 },
|
|
|
|
{ OP_OPEN_PAREN, "(", 0 },
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
enum {
|
|
|
|
FILT_ERR_NONE,
|
|
|
|
FILT_ERR_INVALID_OP,
|
|
|
|
FILT_ERR_UNBALANCED_PAREN,
|
|
|
|
FILT_ERR_TOO_MANY_OPERANDS,
|
|
|
|
FILT_ERR_OPERAND_TOO_LONG,
|
|
|
|
FILT_ERR_FIELD_NOT_FOUND,
|
|
|
|
FILT_ERR_ILLEGAL_FIELD_OP,
|
|
|
|
FILT_ERR_ILLEGAL_INTVAL,
|
|
|
|
FILT_ERR_BAD_SUBSYS_FILTER,
|
|
|
|
FILT_ERR_TOO_MANY_PREDS,
|
|
|
|
FILT_ERR_MISSING_FIELD,
|
|
|
|
FILT_ERR_INVALID_FILTER,
|
|
|
|
};
|
|
|
|
|
|
|
|
static char *err_text[] = {
|
|
|
|
"No error",
|
|
|
|
"Invalid operator",
|
|
|
|
"Unbalanced parens",
|
|
|
|
"Too many operands",
|
|
|
|
"Operand too long",
|
|
|
|
"Field not found",
|
|
|
|
"Illegal operation for field type",
|
|
|
|
"Illegal integer value",
|
|
|
|
"Couldn't find or set field in one of a subsystem's events",
|
|
|
|
"Too many terms in predicate expression",
|
|
|
|
"Missing field name and/or value",
|
|
|
|
"Meaningless filter expression",
|
|
|
|
};
|
|
|
|
|
|
|
|
struct opstack_op {
|
|
|
|
int op;
|
|
|
|
struct list_head list;
|
|
|
|
};
|
|
|
|
|
|
|
|
struct postfix_elt {
|
|
|
|
int op;
|
|
|
|
char *operand;
|
|
|
|
struct list_head list;
|
|
|
|
};
|
|
|
|
|
|
|
|
struct filter_parse_state {
|
|
|
|
struct filter_op *ops;
|
|
|
|
struct list_head opstack;
|
|
|
|
struct list_head postfix;
|
|
|
|
int lasterr;
|
|
|
|
int lasterr_pos;
|
|
|
|
|
|
|
|
struct {
|
|
|
|
char *string;
|
|
|
|
unsigned int cnt;
|
|
|
|
unsigned int tail;
|
|
|
|
} infix;
|
|
|
|
|
|
|
|
struct {
|
|
|
|
char string[MAX_FILTER_STR_VAL];
|
|
|
|
int pos;
|
|
|
|
unsigned int tail;
|
|
|
|
} operand;
|
|
|
|
};
|
|
|
|
|
2009-09-10 03:34:19 +02:00
|
|
|
#define DEFINE_COMPARISON_PRED(type) \
|
|
|
|
static int filter_pred_##type(struct filter_pred *pred, void *event, \
|
|
|
|
int val1, int val2) \
|
|
|
|
{ \
|
|
|
|
type *addr = (type *)(event + pred->offset); \
|
|
|
|
type val = (type)pred->val; \
|
|
|
|
int match = 0; \
|
|
|
|
\
|
|
|
|
switch (pred->op) { \
|
|
|
|
case OP_LT: \
|
|
|
|
match = (*addr < val); \
|
|
|
|
break; \
|
|
|
|
case OP_LE: \
|
|
|
|
match = (*addr <= val); \
|
|
|
|
break; \
|
|
|
|
case OP_GT: \
|
|
|
|
match = (*addr > val); \
|
|
|
|
break; \
|
|
|
|
case OP_GE: \
|
|
|
|
match = (*addr >= val); \
|
|
|
|
break; \
|
|
|
|
default: \
|
|
|
|
break; \
|
|
|
|
} \
|
|
|
|
\
|
|
|
|
return match; \
|
|
|
|
}
|
|
|
|
|
|
|
|
#define DEFINE_EQUALITY_PRED(size) \
|
|
|
|
static int filter_pred_##size(struct filter_pred *pred, void *event, \
|
|
|
|
int val1, int val2) \
|
|
|
|
{ \
|
|
|
|
u##size *addr = (u##size *)(event + pred->offset); \
|
|
|
|
u##size val = (u##size)pred->val; \
|
|
|
|
int match; \
|
|
|
|
\
|
|
|
|
match = (val == *addr) ^ pred->not; \
|
|
|
|
\
|
|
|
|
return match; \
|
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
DEFINE_COMPARISON_PRED(s64);
|
|
|
|
DEFINE_COMPARISON_PRED(u64);
|
|
|
|
DEFINE_COMPARISON_PRED(s32);
|
|
|
|
DEFINE_COMPARISON_PRED(u32);
|
|
|
|
DEFINE_COMPARISON_PRED(s16);
|
|
|
|
DEFINE_COMPARISON_PRED(u16);
|
|
|
|
DEFINE_COMPARISON_PRED(s8);
|
|
|
|
DEFINE_COMPARISON_PRED(u8);
|
|
|
|
|
|
|
|
DEFINE_EQUALITY_PRED(64);
|
|
|
|
DEFINE_EQUALITY_PRED(32);
|
|
|
|
DEFINE_EQUALITY_PRED(16);
|
|
|
|
DEFINE_EQUALITY_PRED(8);
|
|
|
|
|
|
|
|
static int filter_pred_and(struct filter_pred *pred __attribute((unused)),
|
|
|
|
void *event __attribute((unused)),
|
|
|
|
int val1, int val2)
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
return val1 && val2;
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static int filter_pred_or(struct filter_pred *pred __attribute((unused)),
|
|
|
|
void *event __attribute((unused)),
|
|
|
|
int val1, int val2)
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
return val1 || val2;
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
|
|
|
|
2009-05-03 02:48:52 +02:00
|
|
|
/* Filter predicate for fixed sized arrays of characters */
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static int filter_pred_string(struct filter_pred *pred, void *event,
|
|
|
|
int val1, int val2)
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
|
|
|
char *addr = (char *)(event + pred->offset);
|
|
|
|
int cmp, match;
|
|
|
|
|
2009-09-24 21:10:44 +02:00
|
|
|
cmp = pred->regex.match(addr, &pred->regex, pred->regex.field_len);
|
2009-03-22 09:31:04 +01:00
|
|
|
|
2009-09-24 21:10:44 +02:00
|
|
|
match = cmp ^ pred->not;
|
2009-03-22 09:31:04 +01:00
|
|
|
|
|
|
|
return match;
|
|
|
|
}
|
|
|
|
|
2009-08-07 04:33:43 +02:00
|
|
|
/* Filter predicate for char * pointers */
|
|
|
|
static int filter_pred_pchar(struct filter_pred *pred, void *event,
|
|
|
|
int val1, int val2)
|
|
|
|
{
|
|
|
|
char **addr = (char **)(event + pred->offset);
|
|
|
|
int cmp, match;
|
2010-01-14 03:54:27 +01:00
|
|
|
int len = strlen(*addr) + 1; /* including tailing '\0' */
|
2009-08-07 04:33:43 +02:00
|
|
|
|
2010-01-14 03:54:27 +01:00
|
|
|
cmp = pred->regex.match(*addr, &pred->regex, len);
|
2009-08-07 04:33:43 +02:00
|
|
|
|
2009-09-24 21:10:44 +02:00
|
|
|
match = cmp ^ pred->not;
|
2009-08-07 04:33:43 +02:00
|
|
|
|
|
|
|
return match;
|
|
|
|
}
|
|
|
|
|
2009-05-03 02:48:52 +02:00
|
|
|
/*
|
|
|
|
* Filter predicate for dynamic sized arrays of characters.
|
|
|
|
* These are implemented through a list of strings at the end
|
|
|
|
* of the entry.
|
|
|
|
* Also each of these strings have a field in the entry which
|
|
|
|
* contains its offset from the beginning of the entry.
|
|
|
|
* We have then first to get this field, dereference it
|
|
|
|
* and add it to the address of the entry, and at last we have
|
|
|
|
* the address of the string.
|
|
|
|
*/
|
|
|
|
static int filter_pred_strloc(struct filter_pred *pred, void *event,
|
|
|
|
int val1, int val2)
|
|
|
|
{
|
2009-07-16 04:54:02 +02:00
|
|
|
u32 str_item = *(u32 *)(event + pred->offset);
|
|
|
|
int str_loc = str_item & 0xffff;
|
|
|
|
int str_len = str_item >> 16;
|
2009-05-03 02:48:52 +02:00
|
|
|
char *addr = (char *)(event + str_loc);
|
|
|
|
int cmp, match;
|
|
|
|
|
2009-09-24 21:10:44 +02:00
|
|
|
cmp = pred->regex.match(addr, &pred->regex, str_len);
|
2009-05-03 02:48:52 +02:00
|
|
|
|
2009-09-24 21:10:44 +02:00
|
|
|
match = cmp ^ pred->not;
|
2009-05-03 02:48:52 +02:00
|
|
|
|
|
|
|
return match;
|
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static int filter_pred_none(struct filter_pred *pred, void *event,
|
|
|
|
int val1, int val2)
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-01-14 03:54:40 +01:00
|
|
|
/*
|
|
|
|
* regex_match_foo - Basic regex callbacks
|
|
|
|
*
|
|
|
|
* @str: the string to be searched
|
|
|
|
* @r: the regex structure containing the pattern string
|
|
|
|
* @len: the length of the string to be searched (including '\0')
|
|
|
|
*
|
|
|
|
* Note:
|
|
|
|
* - @str might not be NULL-terminated if it's of type DYN_STRING
|
|
|
|
* or STATIC_STRING
|
|
|
|
*/
|
|
|
|
|
2009-09-24 21:10:44 +02:00
|
|
|
static int regex_match_full(char *str, struct regex *r, int len)
|
|
|
|
{
|
|
|
|
if (strncmp(str, r->pattern, len) == 0)
|
|
|
|
return 1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int regex_match_front(char *str, struct regex *r, int len)
|
|
|
|
{
|
2010-01-14 03:53:21 +01:00
|
|
|
if (strncmp(str, r->pattern, r->len) == 0)
|
2009-09-24 21:10:44 +02:00
|
|
|
return 1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int regex_match_middle(char *str, struct regex *r, int len)
|
|
|
|
{
|
2010-01-14 03:54:11 +01:00
|
|
|
if (strnstr(str, r->pattern, len))
|
2009-09-24 21:10:44 +02:00
|
|
|
return 1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int regex_match_end(char *str, struct regex *r, int len)
|
|
|
|
{
|
2010-01-14 03:53:41 +01:00
|
|
|
int strlen = len - 1;
|
2009-09-24 21:10:44 +02:00
|
|
|
|
2010-01-14 03:53:41 +01:00
|
|
|
if (strlen >= r->len &&
|
|
|
|
memcmp(str + strlen - r->len, r->pattern, r->len) == 0)
|
2009-09-24 21:10:44 +02:00
|
|
|
return 1;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-09-24 21:31:51 +02:00
|
|
|
/**
|
|
|
|
* filter_parse_regex - parse a basic regex
|
|
|
|
* @buff: the raw regex
|
|
|
|
* @len: length of the regex
|
|
|
|
* @search: will point to the beginning of the string to compare
|
|
|
|
* @not: tell whether the match will have to be inverted
|
|
|
|
*
|
|
|
|
* This passes in a buffer containing a regex and this function will
|
2009-09-24 21:10:44 +02:00
|
|
|
* set search to point to the search part of the buffer and
|
|
|
|
* return the type of search it is (see enum above).
|
|
|
|
* This does modify buff.
|
|
|
|
*
|
|
|
|
* Returns enum type.
|
|
|
|
* search returns the pointer to use for comparison.
|
|
|
|
* not returns 1 if buff started with a '!'
|
|
|
|
* 0 otherwise.
|
|
|
|
*/
|
2009-09-24 21:31:51 +02:00
|
|
|
enum regex_type filter_parse_regex(char *buff, int len, char **search, int *not)
|
2009-09-24 21:10:44 +02:00
|
|
|
{
|
|
|
|
int type = MATCH_FULL;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (buff[0] == '!') {
|
|
|
|
*not = 1;
|
|
|
|
buff++;
|
|
|
|
len--;
|
|
|
|
} else
|
|
|
|
*not = 0;
|
|
|
|
|
|
|
|
*search = buff;
|
|
|
|
|
|
|
|
for (i = 0; i < len; i++) {
|
|
|
|
if (buff[i] == '*') {
|
|
|
|
if (!i) {
|
|
|
|
*search = buff + 1;
|
|
|
|
type = MATCH_END_ONLY;
|
|
|
|
} else {
|
|
|
|
if (type == MATCH_END_ONLY)
|
|
|
|
type = MATCH_MIDDLE_ONLY;
|
|
|
|
else
|
|
|
|
type = MATCH_FRONT_ONLY;
|
|
|
|
buff[i] = 0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return type;
|
|
|
|
}
|
|
|
|
|
2009-10-15 05:21:12 +02:00
|
|
|
static void filter_build_regex(struct filter_pred *pred)
|
2009-09-24 21:10:44 +02:00
|
|
|
{
|
|
|
|
struct regex *r = &pred->regex;
|
2009-10-15 05:21:12 +02:00
|
|
|
char *search;
|
|
|
|
enum regex_type type = MATCH_FULL;
|
|
|
|
int not = 0;
|
|
|
|
|
|
|
|
if (pred->op == OP_GLOB) {
|
|
|
|
type = filter_parse_regex(r->pattern, r->len, &search, ¬);
|
|
|
|
r->len = strlen(search);
|
|
|
|
memmove(r->pattern, search, r->len+1);
|
|
|
|
}
|
2009-09-24 21:10:44 +02:00
|
|
|
|
|
|
|
switch (type) {
|
|
|
|
case MATCH_FULL:
|
|
|
|
r->match = regex_match_full;
|
|
|
|
break;
|
|
|
|
case MATCH_FRONT_ONLY:
|
|
|
|
r->match = regex_match_front;
|
|
|
|
break;
|
|
|
|
case MATCH_MIDDLE_ONLY:
|
|
|
|
r->match = regex_match_middle;
|
|
|
|
break;
|
|
|
|
case MATCH_END_ONLY:
|
|
|
|
r->match = regex_match_end;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
pred->not ^= not;
|
|
|
|
}
|
|
|
|
|
2009-03-22 09:31:04 +01:00
|
|
|
/* return 1 if event matches, 0 otherwise (discard) */
|
2009-10-15 05:21:42 +02:00
|
|
|
int filter_match_preds(struct event_filter *filter, void *rec)
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
int match, top = 0, val1 = 0, val2 = 0;
|
|
|
|
int stack[MAX_FILTER_PRED];
|
2009-03-22 09:31:04 +01:00
|
|
|
struct filter_pred *pred;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
int i;
|
2009-03-22 09:31:04 +01:00
|
|
|
|
2009-04-28 10:04:47 +02:00
|
|
|
for (i = 0; i < filter->n_preds; i++) {
|
|
|
|
pred = filter->preds[i];
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (!pred->pop_n) {
|
|
|
|
match = pred->fn(pred, rec, val1, val2);
|
|
|
|
stack[top++] = match;
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
continue;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
}
|
|
|
|
if (pred->pop_n > top) {
|
|
|
|
WARN_ON_ONCE(1);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
val1 = stack[--top];
|
|
|
|
val2 = stack[--top];
|
|
|
|
match = pred->fn(pred, rec, val1, val2);
|
|
|
|
stack[top++] = match;
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
return stack[--top];
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
2009-04-11 00:12:50 +02:00
|
|
|
EXPORT_SYMBOL_GPL(filter_match_preds);
|
2009-03-22 09:31:04 +01:00
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static void parse_error(struct filter_parse_state *ps, int err, int pos)
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
ps->lasterr = err;
|
|
|
|
ps->lasterr_pos = pos;
|
|
|
|
}
|
2009-03-22 09:31:04 +01:00
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static void remove_filter_string(struct event_filter *filter)
|
|
|
|
{
|
|
|
|
kfree(filter->filter_string);
|
|
|
|
filter->filter_string = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int replace_filter_string(struct event_filter *filter,
|
|
|
|
char *filter_string)
|
|
|
|
{
|
|
|
|
kfree(filter->filter_string);
|
|
|
|
filter->filter_string = kstrdup(filter_string, GFP_KERNEL);
|
|
|
|
if (!filter->filter_string)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int append_filter_string(struct event_filter *filter,
|
|
|
|
char *string)
|
|
|
|
{
|
|
|
|
int newlen;
|
|
|
|
char *new_filter_string;
|
|
|
|
|
|
|
|
BUG_ON(!filter->filter_string);
|
|
|
|
newlen = strlen(filter->filter_string) + strlen(string) + 1;
|
|
|
|
new_filter_string = kmalloc(newlen, GFP_KERNEL);
|
|
|
|
if (!new_filter_string)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
strcpy(new_filter_string, filter->filter_string);
|
|
|
|
strcat(new_filter_string, string);
|
|
|
|
kfree(filter->filter_string);
|
|
|
|
filter->filter_string = new_filter_string;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void append_filter_err(struct filter_parse_state *ps,
|
|
|
|
struct event_filter *filter)
|
|
|
|
{
|
|
|
|
int pos = ps->lasterr_pos;
|
|
|
|
char *buf, *pbuf;
|
|
|
|
|
|
|
|
buf = (char *)__get_free_page(GFP_TEMPORARY);
|
|
|
|
if (!buf)
|
2009-03-24 08:14:31 +01:00
|
|
|
return;
|
2009-03-22 09:31:04 +01:00
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
append_filter_string(filter, "\n");
|
|
|
|
memset(buf, ' ', PAGE_SIZE);
|
|
|
|
if (pos > PAGE_SIZE - 128)
|
|
|
|
pos = 0;
|
|
|
|
buf[pos] = '^';
|
|
|
|
pbuf = &buf[pos] + 1;
|
|
|
|
|
|
|
|
sprintf(pbuf, "\nparse_error: %s\n", err_text[ps->lasterr]);
|
|
|
|
append_filter_string(filter, buf);
|
|
|
|
free_page((unsigned long) buf);
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
void print_event_filter(struct ftrace_event_call *call, struct trace_seq *s)
|
2009-04-17 07:27:08 +02:00
|
|
|
{
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
struct event_filter *filter = call->filter;
|
|
|
|
|
2009-06-16 10:39:41 +02:00
|
|
|
mutex_lock(&event_mutex);
|
2009-08-31 10:49:41 +02:00
|
|
|
if (filter && filter->filter_string)
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
trace_seq_printf(s, "%s\n", filter->filter_string);
|
|
|
|
else
|
|
|
|
trace_seq_printf(s, "none\n");
|
2009-06-16 10:39:41 +02:00
|
|
|
mutex_unlock(&event_mutex);
|
2009-04-17 07:27:08 +02:00
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
void print_subsystem_event_filter(struct event_subsystem *system,
|
2009-04-17 07:27:08 +02:00
|
|
|
struct trace_seq *s)
|
|
|
|
{
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
struct event_filter *filter = system->filter;
|
|
|
|
|
2009-06-16 10:39:41 +02:00
|
|
|
mutex_lock(&event_mutex);
|
2009-08-31 10:49:41 +02:00
|
|
|
if (filter && filter->filter_string)
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
trace_seq_printf(s, "%s\n", filter->filter_string);
|
|
|
|
else
|
|
|
|
trace_seq_printf(s, "none\n");
|
2009-06-16 10:39:41 +02:00
|
|
|
mutex_unlock(&event_mutex);
|
2009-04-17 07:27:08 +02:00
|
|
|
}
|
|
|
|
|
2009-03-22 09:31:04 +01:00
|
|
|
static struct ftrace_event_field *
|
|
|
|
find_event_field(struct ftrace_event_call *call, char *name)
|
|
|
|
{
|
2009-03-24 08:14:01 +01:00
|
|
|
struct ftrace_event_field *field;
|
2010-04-22 16:35:55 +02:00
|
|
|
struct list_head *head;
|
2009-03-22 09:31:04 +01:00
|
|
|
|
2010-04-22 16:35:55 +02:00
|
|
|
head = trace_get_fields(call);
|
|
|
|
list_for_each_entry(field, head, link) {
|
2009-03-22 09:31:04 +01:00
|
|
|
if (!strcmp(field->name, name))
|
|
|
|
return field;
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static void filter_free_pred(struct filter_pred *pred)
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
|
|
|
if (!pred)
|
|
|
|
return;
|
|
|
|
|
|
|
|
kfree(pred->field_name);
|
|
|
|
kfree(pred);
|
|
|
|
}
|
|
|
|
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
static void filter_clear_pred(struct filter_pred *pred)
|
|
|
|
{
|
|
|
|
kfree(pred->field_name);
|
|
|
|
pred->field_name = NULL;
|
2009-09-24 21:10:44 +02:00
|
|
|
pred->regex.len = 0;
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_set_pred(struct filter_pred *dest,
|
|
|
|
struct filter_pred *src,
|
|
|
|
filter_pred_fn_t fn)
|
|
|
|
{
|
|
|
|
*dest = *src;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (src->field_name) {
|
|
|
|
dest->field_name = kstrdup(src->field_name, GFP_KERNEL);
|
|
|
|
if (!dest->field_name)
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
dest->fn = fn;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static void filter_disable_preds(struct ftrace_event_call *call)
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
2009-04-28 10:04:47 +02:00
|
|
|
struct event_filter *filter = call->filter;
|
2009-03-22 09:31:04 +01:00
|
|
|
int i;
|
|
|
|
|
2010-04-23 17:12:36 +02:00
|
|
|
call->flags &= ~TRACE_EVENT_FL_FILTERED;
|
2009-04-28 10:04:47 +02:00
|
|
|
filter->n_preds = 0;
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
|
|
|
|
for (i = 0; i < MAX_FILTER_PRED; i++)
|
2009-04-28 10:04:47 +02:00
|
|
|
filter->preds[i]->fn = filter_pred_none;
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
}
|
|
|
|
|
2009-10-15 05:21:42 +02:00
|
|
|
static void __free_preds(struct event_filter *filter)
|
2009-05-06 04:33:04 +02:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
2009-08-31 10:49:41 +02:00
|
|
|
if (!filter)
|
|
|
|
return;
|
|
|
|
|
2009-05-06 04:33:04 +02:00
|
|
|
for (i = 0; i < MAX_FILTER_PRED; i++) {
|
|
|
|
if (filter->preds[i])
|
|
|
|
filter_free_pred(filter->preds[i]);
|
|
|
|
}
|
|
|
|
kfree(filter->preds);
|
2009-06-16 10:39:12 +02:00
|
|
|
kfree(filter->filter_string);
|
2009-05-06 04:33:04 +02:00
|
|
|
kfree(filter);
|
2009-10-15 05:21:42 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
void destroy_preds(struct ftrace_event_call *call)
|
|
|
|
{
|
|
|
|
__free_preds(call->filter);
|
2009-05-06 04:33:04 +02:00
|
|
|
call->filter = NULL;
|
2010-04-23 17:12:36 +02:00
|
|
|
call->flags &= ~TRACE_EVENT_FL_FILTERED;
|
2009-05-06 04:33:04 +02:00
|
|
|
}
|
|
|
|
|
2009-10-15 05:21:42 +02:00
|
|
|
static struct event_filter *__alloc_preds(void)
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
{
|
2009-04-28 10:04:47 +02:00
|
|
|
struct event_filter *filter;
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
struct filter_pred *pred;
|
|
|
|
int i;
|
|
|
|
|
2009-10-15 05:21:42 +02:00
|
|
|
filter = kzalloc(sizeof(*filter), GFP_KERNEL);
|
|
|
|
if (!filter)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
|
2009-04-28 10:04:47 +02:00
|
|
|
filter->n_preds = 0;
|
|
|
|
|
|
|
|
filter->preds = kzalloc(MAX_FILTER_PRED * sizeof(pred), GFP_KERNEL);
|
|
|
|
if (!filter->preds)
|
|
|
|
goto oom;
|
|
|
|
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
for (i = 0; i < MAX_FILTER_PRED; i++) {
|
|
|
|
pred = kzalloc(sizeof(*pred), GFP_KERNEL);
|
|
|
|
if (!pred)
|
|
|
|
goto oom;
|
|
|
|
pred->fn = filter_pred_none;
|
2009-04-28 10:04:47 +02:00
|
|
|
filter->preds[i] = pred;
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
}
|
|
|
|
|
2009-10-15 05:21:42 +02:00
|
|
|
return filter;
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
|
|
|
|
oom:
|
2009-10-15 05:21:42 +02:00
|
|
|
__free_preds(filter);
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int init_preds(struct ftrace_event_call *call)
|
|
|
|
{
|
|
|
|
if (call->filter)
|
|
|
|
return 0;
|
|
|
|
|
2010-04-23 17:12:36 +02:00
|
|
|
call->flags &= ~TRACE_EVENT_FL_FILTERED;
|
2009-10-15 05:21:42 +02:00
|
|
|
call->filter = __alloc_preds();
|
|
|
|
if (IS_ERR(call->filter))
|
|
|
|
return PTR_ERR(call->filter);
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
|
2009-10-15 05:21:42 +02:00
|
|
|
return 0;
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
2009-08-31 10:49:41 +02:00
|
|
|
|
|
|
|
static int init_subsystem_preds(struct event_subsystem *system)
|
|
|
|
{
|
|
|
|
struct ftrace_event_call *call;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
list_for_each_entry(call, &ftrace_events, list) {
|
2010-04-22 16:35:55 +02:00
|
|
|
if (!call->class || !call->class->define_fields)
|
2009-08-31 10:49:41 +02:00
|
|
|
continue;
|
|
|
|
|
2010-04-20 16:47:33 +02:00
|
|
|
if (strcmp(call->class->system, system->name) != 0)
|
2009-08-31 10:49:41 +02:00
|
|
|
continue;
|
|
|
|
|
2009-09-01 07:31:38 +02:00
|
|
|
err = init_preds(call);
|
|
|
|
if (err)
|
|
|
|
return err;
|
2009-08-31 10:49:41 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2009-03-22 09:31:04 +01:00
|
|
|
|
2009-10-15 05:20:34 +02:00
|
|
|
static void filter_free_subsystem_preds(struct event_subsystem *system)
|
2009-03-22 09:31:17 +01:00
|
|
|
{
|
2009-04-10 19:52:20 +02:00
|
|
|
struct ftrace_event_call *call;
|
2009-03-22 09:31:17 +01:00
|
|
|
|
2009-04-10 19:52:20 +02:00
|
|
|
list_for_each_entry(call, &ftrace_events, list) {
|
2010-04-22 16:35:55 +02:00
|
|
|
if (!call->class || !call->class->define_fields)
|
2009-03-22 09:31:17 +01:00
|
|
|
continue;
|
|
|
|
|
2010-04-20 16:47:33 +02:00
|
|
|
if (strcmp(call->class->system, system->name) != 0)
|
2009-08-31 10:49:41 +02:00
|
|
|
continue;
|
|
|
|
|
|
|
|
filter_disable_preds(call);
|
|
|
|
remove_filter_string(call->filter);
|
2009-03-22 09:31:17 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static int filter_add_pred_fn(struct filter_parse_state *ps,
|
|
|
|
struct ftrace_event_call *call,
|
2009-10-15 05:21:42 +02:00
|
|
|
struct event_filter *filter,
|
2009-04-17 07:27:08 +02:00
|
|
|
struct filter_pred *pred,
|
|
|
|
filter_pred_fn_t fn)
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
int idx, err;
|
2009-03-22 09:31:04 +01:00
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (filter->n_preds == MAX_FILTER_PRED) {
|
|
|
|
parse_error(ps, FILT_ERR_TOO_MANY_PREDS, 0);
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
return -ENOSPC;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
}
|
2009-03-22 09:31:04 +01:00
|
|
|
|
2009-04-28 10:04:47 +02:00
|
|
|
idx = filter->n_preds;
|
|
|
|
filter_clear_pred(filter->preds[idx]);
|
|
|
|
err = filter_set_pred(filter->preds[idx], pred, fn);
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2009-04-28 10:04:47 +02:00
|
|
|
filter->n_preds++;
|
2009-03-22 09:31:04 +01:00
|
|
|
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
return 0;
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
|
|
|
|
2009-08-07 04:33:02 +02:00
|
|
|
int filter_assign_type(const char *type)
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
tracing/events: introduce __dynamic_array()
__string() is limited:
- it's a char array, but we may want to define array with other types
- a source string should be available, but we may just know the string size
We introduce __dynamic_array() to break those limitations, and __string()
becomes a wrapper of it. As a side effect, now __get_str() can be used
in TP_fast_assign but not only TP_print.
Take XFS for example, we have the string length in the dirent, but the
string itself is not NULL-terminated, so __dynamic_array() can be used:
TRACE_EVENT(xfs_dir2,
TP_PROTO(struct xfs_da_args *args),
TP_ARGS(args),
TP_STRUCT__entry(
__field(int, namelen)
__dynamic_array(char, name, args->namelen + 1)
...
),
TP_fast_assign(
char *name = __get_str(name);
if (args->namelen)
memcpy(name, args->name, args->namelen);
name[args->namelen] = '\0';
__entry->namelen = args->namelen;
),
TP_printk("name %.*s namelen %d",
__entry->namelen ? __get_str(name) : NULL
__entry->namelen)
);
[ Impact: allow defining dynamic size arrays ]
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <4A2384D2.3080403@cn.fujitsu.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2009-06-01 09:35:46 +02:00
|
|
|
if (strstr(type, "__data_loc") && strstr(type, "char"))
|
|
|
|
return FILTER_DYN_STRING;
|
|
|
|
|
2009-03-22 09:31:04 +01:00
|
|
|
if (strchr(type, '[') && strstr(type, "char"))
|
2009-05-03 02:48:52 +02:00
|
|
|
return FILTER_STATIC_STRING;
|
|
|
|
|
2009-08-07 04:33:02 +02:00
|
|
|
return FILTER_OTHER;
|
|
|
|
}
|
|
|
|
|
|
|
|
static bool is_string_field(struct ftrace_event_field *field)
|
|
|
|
{
|
|
|
|
return field->filter_type == FILTER_DYN_STRING ||
|
2009-08-07 04:33:43 +02:00
|
|
|
field->filter_type == FILTER_STATIC_STRING ||
|
|
|
|
field->filter_type == FILTER_PTR_STRING;
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static int is_legal_op(struct ftrace_event_field *field, int op)
|
|
|
|
{
|
2009-10-15 05:21:12 +02:00
|
|
|
if (is_string_field(field) &&
|
|
|
|
(op != OP_EQ && op != OP_NE && op != OP_GLOB))
|
|
|
|
return 0;
|
|
|
|
if (!is_string_field(field) && op == OP_GLOB)
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static filter_pred_fn_t select_comparison_fn(int op, int field_size,
|
|
|
|
int field_is_signed)
|
|
|
|
{
|
|
|
|
filter_pred_fn_t fn = NULL;
|
|
|
|
|
|
|
|
switch (field_size) {
|
|
|
|
case 8:
|
|
|
|
if (op == OP_EQ || op == OP_NE)
|
|
|
|
fn = filter_pred_64;
|
|
|
|
else if (field_is_signed)
|
|
|
|
fn = filter_pred_s64;
|
|
|
|
else
|
|
|
|
fn = filter_pred_u64;
|
|
|
|
break;
|
|
|
|
case 4:
|
|
|
|
if (op == OP_EQ || op == OP_NE)
|
|
|
|
fn = filter_pred_32;
|
|
|
|
else if (field_is_signed)
|
|
|
|
fn = filter_pred_s32;
|
|
|
|
else
|
|
|
|
fn = filter_pred_u32;
|
|
|
|
break;
|
|
|
|
case 2:
|
|
|
|
if (op == OP_EQ || op == OP_NE)
|
|
|
|
fn = filter_pred_16;
|
|
|
|
else if (field_is_signed)
|
|
|
|
fn = filter_pred_s16;
|
|
|
|
else
|
|
|
|
fn = filter_pred_u16;
|
|
|
|
break;
|
|
|
|
case 1:
|
|
|
|
if (op == OP_EQ || op == OP_NE)
|
|
|
|
fn = filter_pred_8;
|
|
|
|
else if (field_is_signed)
|
|
|
|
fn = filter_pred_s8;
|
|
|
|
else
|
|
|
|
fn = filter_pred_u8;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
return fn;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_add_pred(struct filter_parse_state *ps,
|
|
|
|
struct ftrace_event_call *call,
|
2009-10-15 05:21:42 +02:00
|
|
|
struct event_filter *filter,
|
2009-07-20 04:20:53 +02:00
|
|
|
struct filter_pred *pred,
|
|
|
|
bool dry_run)
|
2009-03-22 09:31:04 +01:00
|
|
|
{
|
|
|
|
struct ftrace_event_field *field;
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
filter_pred_fn_t fn;
|
2009-04-21 11:12:11 +02:00
|
|
|
unsigned long long val;
|
2009-06-15 04:58:39 +02:00
|
|
|
int ret;
|
2009-03-22 09:31:04 +01:00
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
pred->fn = filter_pred_none;
|
|
|
|
|
|
|
|
if (pred->op == OP_AND) {
|
|
|
|
pred->pop_n = 2;
|
2009-07-20 04:20:53 +02:00
|
|
|
fn = filter_pred_and;
|
|
|
|
goto add_pred_fn;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
} else if (pred->op == OP_OR) {
|
|
|
|
pred->pop_n = 2;
|
2009-07-20 04:20:53 +02:00
|
|
|
fn = filter_pred_or;
|
|
|
|
goto add_pred_fn;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
}
|
|
|
|
|
2009-03-22 09:31:04 +01:00
|
|
|
field = find_event_field(call, pred->field_name);
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (!field) {
|
|
|
|
parse_error(ps, FILT_ERR_FIELD_NOT_FOUND, 0);
|
2009-03-22 09:31:04 +01:00
|
|
|
return -EINVAL;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
}
|
2009-03-22 09:31:04 +01:00
|
|
|
|
|
|
|
pred->offset = field->offset;
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (!is_legal_op(field, pred->op)) {
|
|
|
|
parse_error(ps, FILT_ERR_ILLEGAL_FIELD_OP, 0);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2009-08-07 04:33:02 +02:00
|
|
|
if (is_string_field(field)) {
|
2009-10-15 05:21:12 +02:00
|
|
|
filter_build_regex(pred);
|
2009-08-07 04:33:43 +02:00
|
|
|
|
2009-09-24 21:10:44 +02:00
|
|
|
if (field->filter_type == FILTER_STATIC_STRING) {
|
2009-05-03 02:48:52 +02:00
|
|
|
fn = filter_pred_string;
|
2009-09-24 21:10:44 +02:00
|
|
|
pred->regex.field_len = field->size;
|
|
|
|
} else if (field->filter_type == FILTER_DYN_STRING)
|
2009-10-15 05:21:12 +02:00
|
|
|
fn = filter_pred_strloc;
|
2010-01-14 03:54:27 +01:00
|
|
|
else
|
2009-08-07 04:33:43 +02:00
|
|
|
fn = filter_pred_pchar;
|
2009-03-24 08:14:42 +01:00
|
|
|
} else {
|
2009-06-15 04:58:39 +02:00
|
|
|
if (field->is_signed)
|
2009-09-24 21:10:44 +02:00
|
|
|
ret = strict_strtoll(pred->regex.pattern, 0, &val);
|
2009-06-15 04:58:39 +02:00
|
|
|
else
|
2009-09-24 21:10:44 +02:00
|
|
|
ret = strict_strtoull(pred->regex.pattern, 0, &val);
|
2009-06-15 04:58:39 +02:00
|
|
|
if (ret) {
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
parse_error(ps, FILT_ERR_ILLEGAL_INTVAL, 0);
|
2009-03-24 08:14:42 +01:00
|
|
|
return -EINVAL;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
}
|
2009-04-21 11:12:11 +02:00
|
|
|
pred->val = val;
|
2009-03-22 09:31:04 +01:00
|
|
|
|
2009-07-20 04:20:53 +02:00
|
|
|
fn = select_comparison_fn(pred->op, field->size,
|
|
|
|
field->is_signed);
|
|
|
|
if (!fn) {
|
|
|
|
parse_error(ps, FILT_ERR_INVALID_OP, 0);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (pred->op == OP_NE)
|
|
|
|
pred->not = 1;
|
2009-04-17 07:27:08 +02:00
|
|
|
|
2009-07-20 04:20:53 +02:00
|
|
|
add_pred_fn:
|
|
|
|
if (!dry_run)
|
2009-10-15 05:21:42 +02:00
|
|
|
return filter_add_pred_fn(ps, call, filter, pred, fn);
|
2009-07-20 04:20:53 +02:00
|
|
|
return 0;
|
2009-03-22 09:31:17 +01:00
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static void parse_init(struct filter_parse_state *ps,
|
|
|
|
struct filter_op *ops,
|
|
|
|
char *infix_string)
|
|
|
|
{
|
|
|
|
memset(ps, '\0', sizeof(*ps));
|
|
|
|
|
|
|
|
ps->infix.string = infix_string;
|
|
|
|
ps->infix.cnt = strlen(infix_string);
|
|
|
|
ps->ops = ops;
|
|
|
|
|
|
|
|
INIT_LIST_HEAD(&ps->opstack);
|
|
|
|
INIT_LIST_HEAD(&ps->postfix);
|
|
|
|
}
|
|
|
|
|
|
|
|
static char infix_next(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
ps->infix.cnt--;
|
|
|
|
|
|
|
|
return ps->infix.string[ps->infix.tail++];
|
|
|
|
}
|
|
|
|
|
|
|
|
static char infix_peek(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
if (ps->infix.tail == strlen(ps->infix.string))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
return ps->infix.string[ps->infix.tail];
|
|
|
|
}
|
|
|
|
|
|
|
|
static void infix_advance(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
ps->infix.cnt--;
|
|
|
|
ps->infix.tail++;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int is_precedence_lower(struct filter_parse_state *ps,
|
|
|
|
int a, int b)
|
|
|
|
{
|
|
|
|
return ps->ops[a].precedence < ps->ops[b].precedence;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int is_op_char(struct filter_parse_state *ps, char c)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; strcmp(ps->ops[i].string, "OP_NONE"); i++) {
|
|
|
|
if (ps->ops[i].string[0] == c)
|
|
|
|
return 1;
|
|
|
|
}
|
2009-03-23 09:26:48 +01:00
|
|
|
|
tracing/filters: allow on-the-fly filter switching
This patch allows event filters to be safely removed or switched
on-the-fly while avoiding the use of rcu or the suspension of tracing of
previous versions.
It does it by adding a new filter_pred_none() predicate function which
does nothing and by never deallocating either the predicates or any of
the filter_pred members used in matching; the predicate lists are
allocated and initialized during ftrace_event_calls initialization.
Whenever a filter is removed or replaced, the filter_pred_* functions
currently in use by the affected ftrace_event_call are immediately
switched over to to the filter_pred_none() function, while the rest of
the filter_pred members are left intact, allowing any currently
executing filter_pred_* functions to finish up, using the values they're
currently using.
In the case of filter replacement, the new predicate values are copied
into the old predicates after the above step, and the filter_pred_none()
functions are replaced by the filter_pred_* functions for the new
filter. In this case, it is possible though very unlikely that a
previous filter_pred_* is still running even after the
filter_pred_none() switch and the switch to the new filter_pred_*. In
that case, however, because nothing has been deallocated in the
filter_pred, the worst that can happen is that the old filter_pred_*
function sees the new values and as a result produces either a false
positive or a false negative, depending on the values it finds.
So one downside to this method is that rarely, it can produce a bad
match during the filter switch, but it should be possible to live with
that, IMHO.
The other downside is that at least in this patch the predicate lists
are always pre-allocated, taking up memory from the start. They could
probably be allocated on first-use, and de-allocated when tracing is
completely stopped - if this patch makes sense, I could create another
one to do that later on.
Oh, and it also places a restriction on the size of __arrays in events,
currently set to 128, since they can't be larger than the now embedded
str_val arrays in the filter_pred struct.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: paulmck@linux.vnet.ibm.com
LKML-Reference: <1239610670.6660.49.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-13 10:17:50 +02:00
|
|
|
return 0;
|
2009-03-22 09:31:17 +01:00
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
static int infix_get_op(struct filter_parse_state *ps, char firstc)
|
|
|
|
{
|
|
|
|
char nextc = infix_peek(ps);
|
|
|
|
char opstr[3];
|
|
|
|
int i;
|
|
|
|
|
|
|
|
opstr[0] = firstc;
|
|
|
|
opstr[1] = nextc;
|
|
|
|
opstr[2] = '\0';
|
|
|
|
|
|
|
|
for (i = 0; strcmp(ps->ops[i].string, "OP_NONE"); i++) {
|
|
|
|
if (!strcmp(opstr, ps->ops[i].string)) {
|
|
|
|
infix_advance(ps);
|
|
|
|
return ps->ops[i].id;
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
opstr[1] = '\0';
|
|
|
|
|
|
|
|
for (i = 0; strcmp(ps->ops[i].string, "OP_NONE"); i++) {
|
|
|
|
if (!strcmp(opstr, ps->ops[i].string))
|
|
|
|
return ps->ops[i].id;
|
|
|
|
}
|
|
|
|
|
|
|
|
return OP_NONE;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void clear_operand_string(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
memset(ps->operand.string, '\0', MAX_FILTER_STR_VAL);
|
|
|
|
ps->operand.tail = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int append_operand_char(struct filter_parse_state *ps, char c)
|
|
|
|
{
|
2009-05-15 05:07:56 +02:00
|
|
|
if (ps->operand.tail == MAX_FILTER_STR_VAL - 1)
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
ps->operand.string[ps->operand.tail++] = c;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_opstack_push(struct filter_parse_state *ps, int op)
|
|
|
|
{
|
|
|
|
struct opstack_op *opstack_op;
|
|
|
|
|
|
|
|
opstack_op = kmalloc(sizeof(*opstack_op), GFP_KERNEL);
|
|
|
|
if (!opstack_op)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
opstack_op->op = op;
|
|
|
|
list_add(&opstack_op->list, &ps->opstack);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_opstack_empty(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
return list_empty(&ps->opstack);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_opstack_top(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
struct opstack_op *opstack_op;
|
|
|
|
|
|
|
|
if (filter_opstack_empty(ps))
|
|
|
|
return OP_NONE;
|
|
|
|
|
|
|
|
opstack_op = list_first_entry(&ps->opstack, struct opstack_op, list);
|
|
|
|
|
|
|
|
return opstack_op->op;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_opstack_pop(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
struct opstack_op *opstack_op;
|
|
|
|
int op;
|
|
|
|
|
|
|
|
if (filter_opstack_empty(ps))
|
|
|
|
return OP_NONE;
|
|
|
|
|
|
|
|
opstack_op = list_first_entry(&ps->opstack, struct opstack_op, list);
|
|
|
|
op = opstack_op->op;
|
|
|
|
list_del(&opstack_op->list);
|
|
|
|
|
|
|
|
kfree(opstack_op);
|
|
|
|
|
|
|
|
return op;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void filter_opstack_clear(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
while (!filter_opstack_empty(ps))
|
|
|
|
filter_opstack_pop(ps);
|
|
|
|
}
|
|
|
|
|
|
|
|
static char *curr_operand(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
return ps->operand.string;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int postfix_append_operand(struct filter_parse_state *ps, char *operand)
|
|
|
|
{
|
|
|
|
struct postfix_elt *elt;
|
|
|
|
|
|
|
|
elt = kmalloc(sizeof(*elt), GFP_KERNEL);
|
|
|
|
if (!elt)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
elt->op = OP_NONE;
|
|
|
|
elt->operand = kstrdup(operand, GFP_KERNEL);
|
|
|
|
if (!elt->operand) {
|
|
|
|
kfree(elt);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
list_add_tail(&elt->list, &ps->postfix);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int postfix_append_op(struct filter_parse_state *ps, int op)
|
|
|
|
{
|
|
|
|
struct postfix_elt *elt;
|
|
|
|
|
|
|
|
elt = kmalloc(sizeof(*elt), GFP_KERNEL);
|
|
|
|
if (!elt)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
elt->op = op;
|
|
|
|
elt->operand = NULL;
|
|
|
|
|
|
|
|
list_add_tail(&elt->list, &ps->postfix);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void postfix_clear(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
struct postfix_elt *elt;
|
|
|
|
|
|
|
|
while (!list_empty(&ps->postfix)) {
|
|
|
|
elt = list_first_entry(&ps->postfix, struct postfix_elt, list);
|
|
|
|
list_del(&elt->list);
|
2009-10-13 03:28:57 +02:00
|
|
|
kfree(elt->operand);
|
|
|
|
kfree(elt);
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_parse(struct filter_parse_state *ps)
|
|
|
|
{
|
2009-05-03 03:03:57 +02:00
|
|
|
int in_string = 0;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
int op, top_op;
|
|
|
|
char ch;
|
|
|
|
|
|
|
|
while ((ch = infix_next(ps))) {
|
2009-05-03 03:03:57 +02:00
|
|
|
if (ch == '"') {
|
|
|
|
in_string ^= 1;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (in_string)
|
|
|
|
goto parse_operand;
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (isspace(ch))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (is_op_char(ps, ch)) {
|
|
|
|
op = infix_get_op(ps, ch);
|
|
|
|
if (op == OP_NONE) {
|
|
|
|
parse_error(ps, FILT_ERR_INVALID_OP, 0);
|
2009-03-22 09:31:04 +01:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
|
|
|
|
if (strlen(curr_operand(ps))) {
|
|
|
|
postfix_append_operand(ps, curr_operand(ps));
|
|
|
|
clear_operand_string(ps);
|
|
|
|
}
|
|
|
|
|
|
|
|
while (!filter_opstack_empty(ps)) {
|
|
|
|
top_op = filter_opstack_top(ps);
|
|
|
|
if (!is_precedence_lower(ps, top_op, op)) {
|
|
|
|
top_op = filter_opstack_pop(ps);
|
|
|
|
postfix_append_op(ps, top_op);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
filter_opstack_push(ps, op);
|
2009-03-22 09:31:04 +01:00
|
|
|
continue;
|
|
|
|
}
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
|
|
|
|
if (ch == '(') {
|
|
|
|
filter_opstack_push(ps, OP_OPEN_PAREN);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (ch == ')') {
|
|
|
|
if (strlen(curr_operand(ps))) {
|
|
|
|
postfix_append_operand(ps, curr_operand(ps));
|
|
|
|
clear_operand_string(ps);
|
|
|
|
}
|
|
|
|
|
|
|
|
top_op = filter_opstack_pop(ps);
|
|
|
|
while (top_op != OP_NONE) {
|
|
|
|
if (top_op == OP_OPEN_PAREN)
|
|
|
|
break;
|
|
|
|
postfix_append_op(ps, top_op);
|
|
|
|
top_op = filter_opstack_pop(ps);
|
|
|
|
}
|
|
|
|
if (top_op == OP_NONE) {
|
|
|
|
parse_error(ps, FILT_ERR_UNBALANCED_PAREN, 0);
|
|
|
|
return -EINVAL;
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
|
|
|
continue;
|
|
|
|
}
|
2009-05-03 03:03:57 +02:00
|
|
|
parse_operand:
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (append_operand_char(ps, ch)) {
|
|
|
|
parse_error(ps, FILT_ERR_OPERAND_TOO_LONG, 0);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (strlen(curr_operand(ps)))
|
|
|
|
postfix_append_operand(ps, curr_operand(ps));
|
|
|
|
|
|
|
|
while (!filter_opstack_empty(ps)) {
|
|
|
|
top_op = filter_opstack_pop(ps);
|
|
|
|
if (top_op == OP_NONE)
|
|
|
|
break;
|
|
|
|
if (top_op == OP_OPEN_PAREN) {
|
|
|
|
parse_error(ps, FILT_ERR_UNBALANCED_PAREN, 0);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
postfix_append_op(ps, top_op);
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct filter_pred *create_pred(int op, char *operand1, char *operand2)
|
|
|
|
{
|
|
|
|
struct filter_pred *pred;
|
|
|
|
|
|
|
|
pred = kzalloc(sizeof(*pred), GFP_KERNEL);
|
|
|
|
if (!pred)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
pred->field_name = kstrdup(operand1, GFP_KERNEL);
|
|
|
|
if (!pred->field_name) {
|
|
|
|
kfree(pred);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2009-09-24 21:10:44 +02:00
|
|
|
strcpy(pred->regex.pattern, operand2);
|
|
|
|
pred->regex.len = strlen(pred->regex.pattern);
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
|
|
|
|
pred->op = op;
|
|
|
|
|
|
|
|
return pred;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct filter_pred *create_logical_pred(int op)
|
|
|
|
{
|
|
|
|
struct filter_pred *pred;
|
|
|
|
|
|
|
|
pred = kzalloc(sizeof(*pred), GFP_KERNEL);
|
|
|
|
if (!pred)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
pred->op = op;
|
|
|
|
|
|
|
|
return pred;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int check_preds(struct filter_parse_state *ps)
|
|
|
|
{
|
|
|
|
int n_normal_preds = 0, n_logical_preds = 0;
|
|
|
|
struct postfix_elt *elt;
|
|
|
|
|
|
|
|
list_for_each_entry(elt, &ps->postfix, list) {
|
|
|
|
if (elt->op == OP_NONE)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (elt->op == OP_AND || elt->op == OP_OR) {
|
|
|
|
n_logical_preds++;
|
|
|
|
continue;
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
n_normal_preds++;
|
2009-03-22 09:31:04 +01:00
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (!n_normal_preds || n_logical_preds >= n_normal_preds) {
|
|
|
|
parse_error(ps, FILT_ERR_INVALID_FILTER, 0);
|
2009-04-11 09:52:35 +02:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
return 0;
|
|
|
|
}
|
2009-04-21 11:12:11 +02:00
|
|
|
|
2009-10-15 05:20:34 +02:00
|
|
|
static int replace_preds(struct ftrace_event_call *call,
|
2009-10-15 05:21:42 +02:00
|
|
|
struct event_filter *filter,
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
struct filter_parse_state *ps,
|
2009-07-20 04:20:53 +02:00
|
|
|
char *filter_string,
|
|
|
|
bool dry_run)
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
{
|
|
|
|
char *operand1 = NULL, *operand2 = NULL;
|
|
|
|
struct filter_pred *pred;
|
|
|
|
struct postfix_elt *elt;
|
|
|
|
int err;
|
2009-07-20 04:20:53 +02:00
|
|
|
int n_preds = 0;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
|
|
|
|
err = check_preds(ps);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
list_for_each_entry(elt, &ps->postfix, list) {
|
|
|
|
if (elt->op == OP_NONE) {
|
|
|
|
if (!operand1)
|
|
|
|
operand1 = elt->operand;
|
|
|
|
else if (!operand2)
|
|
|
|
operand2 = elt->operand;
|
|
|
|
else {
|
|
|
|
parse_error(ps, FILT_ERR_TOO_MANY_OPERANDS, 0);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2009-07-20 04:20:53 +02:00
|
|
|
if (n_preds++ == MAX_FILTER_PRED) {
|
|
|
|
parse_error(ps, FILT_ERR_TOO_MANY_PREDS, 0);
|
|
|
|
return -ENOSPC;
|
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (elt->op == OP_AND || elt->op == OP_OR) {
|
|
|
|
pred = create_logical_pred(elt->op);
|
2009-07-20 04:20:53 +02:00
|
|
|
goto add_pred;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!operand1 || !operand2) {
|
|
|
|
parse_error(ps, FILT_ERR_MISSING_FIELD, 0);
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
pred = create_pred(elt->op, operand1, operand2);
|
2009-07-20 04:20:53 +02:00
|
|
|
add_pred:
|
2009-08-08 17:49:36 +02:00
|
|
|
if (!pred)
|
|
|
|
return -ENOMEM;
|
2009-10-15 05:21:42 +02:00
|
|
|
err = filter_add_pred(ps, call, filter, pred, dry_run);
|
2009-07-09 10:20:12 +02:00
|
|
|
filter_free_pred(pred);
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
operand1 = operand2 = NULL;
|
|
|
|
}
|
2009-03-22 09:31:04 +01:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-10-15 05:20:34 +02:00
|
|
|
static int replace_system_preds(struct event_subsystem *system,
|
|
|
|
struct filter_parse_state *ps,
|
|
|
|
char *filter_string)
|
|
|
|
{
|
|
|
|
struct ftrace_event_call *call;
|
|
|
|
bool fail = true;
|
2009-10-15 12:24:04 +02:00
|
|
|
int err;
|
2009-10-15 05:20:34 +02:00
|
|
|
|
|
|
|
list_for_each_entry(call, &ftrace_events, list) {
|
2009-10-28 10:37:01 +01:00
|
|
|
struct event_filter *filter = call->filter;
|
2009-10-15 05:20:34 +02:00
|
|
|
|
2010-04-22 16:35:55 +02:00
|
|
|
if (!call->class || !call->class->define_fields)
|
2009-10-15 05:20:34 +02:00
|
|
|
continue;
|
|
|
|
|
2010-04-20 16:47:33 +02:00
|
|
|
if (strcmp(call->class->system, system->name) != 0)
|
2009-10-15 05:20:34 +02:00
|
|
|
continue;
|
|
|
|
|
|
|
|
/* try to see if the filter can be applied */
|
2009-10-15 05:21:42 +02:00
|
|
|
err = replace_preds(call, filter, ps, filter_string, true);
|
2009-10-15 05:20:34 +02:00
|
|
|
if (err)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/* really apply the filter */
|
|
|
|
filter_disable_preds(call);
|
2009-10-15 05:21:42 +02:00
|
|
|
err = replace_preds(call, filter, ps, filter_string, false);
|
2009-10-15 05:20:34 +02:00
|
|
|
if (err)
|
|
|
|
filter_disable_preds(call);
|
2009-10-15 05:21:42 +02:00
|
|
|
else {
|
2010-04-23 17:12:36 +02:00
|
|
|
call->flags |= TRACE_EVENT_FL_FILTERED;
|
2009-10-15 05:21:42 +02:00
|
|
|
replace_filter_string(filter, filter_string);
|
|
|
|
}
|
2009-10-15 05:20:34 +02:00
|
|
|
fail = false;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (fail) {
|
|
|
|
parse_error(ps, FILT_ERR_BAD_SUBSYS_FILTER, 0);
|
2009-10-15 12:24:04 +02:00
|
|
|
return -EINVAL;
|
2009-10-15 05:20:34 +02:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
int apply_event_filter(struct ftrace_event_call *call, char *filter_string)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
struct filter_parse_state *ps;
|
|
|
|
|
2009-06-16 10:39:41 +02:00
|
|
|
mutex_lock(&event_mutex);
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
|
2009-08-31 10:49:41 +02:00
|
|
|
err = init_preds(call);
|
|
|
|
if (err)
|
|
|
|
goto out_unlock;
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (!strcmp(strstrip(filter_string), "0")) {
|
|
|
|
filter_disable_preds(call);
|
|
|
|
remove_filter_string(call->filter);
|
2009-10-15 12:24:04 +02:00
|
|
|
goto out_unlock;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
}
|
|
|
|
|
2009-05-15 05:07:27 +02:00
|
|
|
err = -ENOMEM;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
ps = kzalloc(sizeof(*ps), GFP_KERNEL);
|
|
|
|
if (!ps)
|
2009-05-15 05:07:27 +02:00
|
|
|
goto out_unlock;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
|
|
|
|
filter_disable_preds(call);
|
|
|
|
replace_filter_string(call->filter, filter_string);
|
|
|
|
|
|
|
|
parse_init(ps, filter_ops, filter_string);
|
|
|
|
err = filter_parse(ps);
|
|
|
|
if (err) {
|
|
|
|
append_filter_err(ps, call->filter);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2009-10-15 05:21:42 +02:00
|
|
|
err = replace_preds(call, call->filter, ps, filter_string, false);
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (err)
|
|
|
|
append_filter_err(ps, call->filter);
|
2009-10-15 05:21:42 +02:00
|
|
|
else
|
2010-04-23 17:12:36 +02:00
|
|
|
call->flags |= TRACE_EVENT_FL_FILTERED;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
out:
|
|
|
|
filter_opstack_clear(ps);
|
|
|
|
postfix_clear(ps);
|
|
|
|
kfree(ps);
|
2009-05-15 05:07:27 +02:00
|
|
|
out_unlock:
|
2009-06-16 10:39:41 +02:00
|
|
|
mutex_unlock(&event_mutex);
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
int apply_subsystem_event_filter(struct event_subsystem *system,
|
|
|
|
char *filter_string)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
struct filter_parse_state *ps;
|
|
|
|
|
2009-06-16 10:39:41 +02:00
|
|
|
mutex_lock(&event_mutex);
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
|
2009-08-31 10:49:41 +02:00
|
|
|
err = init_subsystem_preds(system);
|
|
|
|
if (err)
|
|
|
|
goto out_unlock;
|
|
|
|
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
if (!strcmp(strstrip(filter_string), "0")) {
|
2009-10-15 05:20:34 +02:00
|
|
|
filter_free_subsystem_preds(system);
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
remove_filter_string(system->filter);
|
2009-10-15 12:24:04 +02:00
|
|
|
goto out_unlock;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
}
|
|
|
|
|
2009-05-15 05:07:27 +02:00
|
|
|
err = -ENOMEM;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
ps = kzalloc(sizeof(*ps), GFP_KERNEL);
|
|
|
|
if (!ps)
|
2009-05-15 05:07:27 +02:00
|
|
|
goto out_unlock;
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
|
|
|
|
replace_filter_string(system->filter, filter_string);
|
|
|
|
|
|
|
|
parse_init(ps, filter_ops, filter_string);
|
|
|
|
err = filter_parse(ps);
|
|
|
|
if (err) {
|
|
|
|
append_filter_err(ps, system->filter);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2009-10-15 05:20:34 +02:00
|
|
|
err = replace_system_preds(system, ps, filter_string);
|
|
|
|
if (err)
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
append_filter_err(ps, system->filter);
|
|
|
|
|
|
|
|
out:
|
|
|
|
filter_opstack_clear(ps);
|
|
|
|
postfix_clear(ps);
|
|
|
|
kfree(ps);
|
2009-05-15 05:07:27 +02:00
|
|
|
out_unlock:
|
2009-06-16 10:39:41 +02:00
|
|
|
mutex_unlock(&event_mutex);
|
tracing/filters: a better event parser
Replace the current event parser hack with a better one. Filters are
no longer specified predicate by predicate, but all at once and can
use parens and any of the following operators:
numeric fields:
==, !=, <, <=, >, >=
string fields:
==, !=
predicates can be combined with the logical operators:
&&, ||
examples:
"common_preempt_count > 4" > filter
"((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
If there was an error, the erroneous string along with an error
message can be seen by looking at the filter e.g.:
((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
^
parse_error: Field not found
Currently the caret for an error always appears at the beginning of
the filter; a real position should be used, but the error message
should be useful even without it.
To clear a filter, '0' can be written to the filter file.
Filters can also be set or cleared for a complete subsystem by writing
the same filter as would be written to an individual event to the
filter file at the root of the subsytem. Note however, that if any
event in the subsystem lacks a field specified in the filter being
set, the set will fail and all filters in the subsytem are
automatically cleared. This change from the previous version was made
because using only the fields that happen to exist for a given event
would most likely result in a meaningless filter.
Because the logical operators are now implemented as predicates, the
maximum number of predicates in a filter was increased from 8 to 16.
[ Impact: add new, extended trace-filter implementation ]
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: fweisbec@gmail.com
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1240905899.6416.121.camel@tropicana>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 10:04:59 +02:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
2009-03-22 09:31:04 +01:00
|
|
|
|
2009-12-21 07:27:35 +01:00
|
|
|
#ifdef CONFIG_PERF_EVENTS
|
2009-10-15 05:21:42 +02:00
|
|
|
|
|
|
|
void ftrace_profile_free_filter(struct perf_event *event)
|
|
|
|
{
|
|
|
|
struct event_filter *filter = event->filter;
|
|
|
|
|
|
|
|
event->filter = NULL;
|
|
|
|
__free_preds(filter);
|
|
|
|
}
|
|
|
|
|
|
|
|
int ftrace_profile_set_filter(struct perf_event *event, int event_id,
|
|
|
|
char *filter_str)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
struct event_filter *filter;
|
|
|
|
struct filter_parse_state *ps;
|
|
|
|
struct ftrace_event_call *call = NULL;
|
|
|
|
|
|
|
|
mutex_lock(&event_mutex);
|
|
|
|
|
|
|
|
list_for_each_entry(call, &ftrace_events, list) {
|
2010-04-23 16:38:03 +02:00
|
|
|
if (call->event.type == event_id)
|
2009-10-15 05:21:42 +02:00
|
|
|
break;
|
|
|
|
}
|
2009-10-15 12:24:04 +02:00
|
|
|
|
|
|
|
err = -EINVAL;
|
2010-03-20 15:39:11 +01:00
|
|
|
if (&call->list == &ftrace_events)
|
2009-10-15 12:24:04 +02:00
|
|
|
goto out_unlock;
|
2009-10-15 05:21:42 +02:00
|
|
|
|
2009-10-15 12:24:04 +02:00
|
|
|
err = -EEXIST;
|
2009-10-15 05:21:42 +02:00
|
|
|
if (event->filter)
|
2009-10-15 12:24:04 +02:00
|
|
|
goto out_unlock;
|
2009-10-15 05:21:42 +02:00
|
|
|
|
|
|
|
filter = __alloc_preds();
|
2009-10-15 12:24:04 +02:00
|
|
|
if (IS_ERR(filter)) {
|
|
|
|
err = PTR_ERR(filter);
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
2009-10-15 05:21:42 +02:00
|
|
|
|
|
|
|
err = -ENOMEM;
|
|
|
|
ps = kzalloc(sizeof(*ps), GFP_KERNEL);
|
|
|
|
if (!ps)
|
|
|
|
goto free_preds;
|
|
|
|
|
|
|
|
parse_init(ps, filter_ops, filter_str);
|
|
|
|
err = filter_parse(ps);
|
|
|
|
if (err)
|
|
|
|
goto free_ps;
|
|
|
|
|
|
|
|
err = replace_preds(call, filter, ps, filter_str, false);
|
|
|
|
if (!err)
|
|
|
|
event->filter = filter;
|
|
|
|
|
|
|
|
free_ps:
|
|
|
|
filter_opstack_clear(ps);
|
|
|
|
postfix_clear(ps);
|
|
|
|
kfree(ps);
|
|
|
|
|
|
|
|
free_preds:
|
|
|
|
if (err)
|
|
|
|
__free_preds(filter);
|
|
|
|
|
2009-10-15 12:24:04 +02:00
|
|
|
out_unlock:
|
2009-10-15 05:21:42 +02:00
|
|
|
mutex_unlock(&event_mutex);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2009-12-21 07:27:35 +01:00
|
|
|
#endif /* CONFIG_PERF_EVENTS */
|
2009-10-15 05:21:42 +02:00
|
|
|
|