2009-03-22 08:31:04 +00:00
|
|
|
/*
|
|
|
|
* 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/debugfs.h>
|
|
|
|
#include <linux/uaccess.h>
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/ctype.h>
|
2009-04-17 05:27:08 +00:00
|
|
|
#include <linux/mutex.h>
|
2009-03-22 08:31:04 +00:00
|
|
|
|
|
|
|
#include "trace.h"
|
2009-03-24 07:14:31 +00:00
|
|
|
#include "trace_output.h"
|
2009-03-22 08:31:04 +00:00
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
static DEFINE_MUTEX(filter_mutex);
|
|
|
|
|
2009-03-22 08:31:04 +00:00
|
|
|
static int filter_pred_64(struct filter_pred *pred, void *event)
|
|
|
|
{
|
|
|
|
u64 *addr = (u64 *)(event + pred->offset);
|
|
|
|
u64 val = (u64)pred->val;
|
|
|
|
int match;
|
|
|
|
|
|
|
|
match = (val == *addr) ^ pred->not;
|
|
|
|
|
|
|
|
return match;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_pred_32(struct filter_pred *pred, void *event)
|
|
|
|
{
|
|
|
|
u32 *addr = (u32 *)(event + pred->offset);
|
|
|
|
u32 val = (u32)pred->val;
|
|
|
|
int match;
|
|
|
|
|
|
|
|
match = (val == *addr) ^ pred->not;
|
|
|
|
|
|
|
|
return match;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_pred_16(struct filter_pred *pred, void *event)
|
|
|
|
{
|
|
|
|
u16 *addr = (u16 *)(event + pred->offset);
|
|
|
|
u16 val = (u16)pred->val;
|
|
|
|
int match;
|
|
|
|
|
|
|
|
match = (val == *addr) ^ pred->not;
|
|
|
|
|
|
|
|
return match;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_pred_8(struct filter_pred *pred, void *event)
|
|
|
|
{
|
|
|
|
u8 *addr = (u8 *)(event + pred->offset);
|
|
|
|
u8 val = (u8)pred->val;
|
|
|
|
int match;
|
|
|
|
|
|
|
|
match = (val == *addr) ^ pred->not;
|
|
|
|
|
|
|
|
return match;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_pred_string(struct filter_pred *pred, void *event)
|
|
|
|
{
|
|
|
|
char *addr = (char *)(event + pred->offset);
|
|
|
|
int cmp, match;
|
|
|
|
|
|
|
|
cmp = strncmp(addr, pred->str_val, pred->str_len);
|
|
|
|
|
|
|
|
match = (!cmp) ^ pred->not;
|
|
|
|
|
|
|
|
return 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 08:17:50 +00:00
|
|
|
static int filter_pred_none(struct filter_pred *pred, void *event)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-03-22 08:31:04 +00:00
|
|
|
/* return 1 if event matches, 0 otherwise (discard) */
|
|
|
|
int filter_match_preds(struct ftrace_event_call *call, void *rec)
|
|
|
|
{
|
|
|
|
int i, matched, and_failed = 0;
|
|
|
|
struct filter_pred *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 08:17:50 +00:00
|
|
|
for (i = 0; i < call->n_preds; i++) {
|
|
|
|
pred = call->preds[i];
|
|
|
|
if (and_failed && !pred->or)
|
|
|
|
continue;
|
|
|
|
matched = pred->fn(pred, rec);
|
|
|
|
if (!matched && !pred->or) {
|
|
|
|
and_failed = 1;
|
|
|
|
continue;
|
|
|
|
} else if (matched && pred->or)
|
|
|
|
return 1;
|
2009-03-22 08:31:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (and_failed)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
2009-04-10 22:12:50 +00:00
|
|
|
EXPORT_SYMBOL_GPL(filter_match_preds);
|
2009-03-22 08:31:04 +00:00
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
static void __filter_print_preds(struct filter_pred **preds, int n_preds,
|
|
|
|
struct trace_seq *s)
|
2009-03-22 08:31:04 +00:00
|
|
|
{
|
|
|
|
char *field_name;
|
|
|
|
struct filter_pred *pred;
|
|
|
|
int i;
|
|
|
|
|
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 08:17:50 +00:00
|
|
|
if (!n_preds) {
|
2009-03-24 07:14:31 +00:00
|
|
|
trace_seq_printf(s, "none\n");
|
|
|
|
return;
|
2009-03-22 08:31:04 +00:00
|
|
|
}
|
|
|
|
|
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 08:17:50 +00:00
|
|
|
for (i = 0; i < n_preds; i++) {
|
|
|
|
pred = preds[i];
|
|
|
|
field_name = pred->field_name;
|
|
|
|
if (i)
|
|
|
|
trace_seq_printf(s, pred->or ? "|| " : "&& ");
|
|
|
|
trace_seq_printf(s, "%s ", field_name);
|
|
|
|
trace_seq_printf(s, pred->not ? "!= " : "== ");
|
|
|
|
if (pred->str_len)
|
|
|
|
trace_seq_printf(s, "%s\n", pred->str_val);
|
|
|
|
else
|
|
|
|
trace_seq_printf(s, "%llu\n", pred->val);
|
2009-03-22 08:31:04 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
void filter_print_preds(struct ftrace_event_call *call, struct trace_seq *s)
|
|
|
|
{
|
|
|
|
mutex_lock(&filter_mutex);
|
|
|
|
__filter_print_preds(call->preds, call->n_preds, s);
|
|
|
|
mutex_unlock(&filter_mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
void filter_print_subsystem_preds(struct event_subsystem *system,
|
|
|
|
struct trace_seq *s)
|
|
|
|
{
|
|
|
|
mutex_lock(&filter_mutex);
|
|
|
|
__filter_print_preds(system->preds, system->n_preds, s);
|
|
|
|
mutex_unlock(&filter_mutex);
|
|
|
|
}
|
|
|
|
|
2009-03-22 08:31:04 +00:00
|
|
|
static struct ftrace_event_field *
|
|
|
|
find_event_field(struct ftrace_event_call *call, char *name)
|
|
|
|
{
|
2009-03-24 07:14:01 +00:00
|
|
|
struct ftrace_event_field *field;
|
2009-03-22 08:31:04 +00:00
|
|
|
|
2009-03-24 07:14:01 +00:00
|
|
|
list_for_each_entry(field, &call->fields, link) {
|
2009-03-22 08:31:04 +00:00
|
|
|
if (!strcmp(field->name, name))
|
|
|
|
return field;
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
void filter_free_pred(struct filter_pred *pred)
|
|
|
|
{
|
|
|
|
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 08:17:50 +00:00
|
|
|
static void filter_clear_pred(struct filter_pred *pred)
|
|
|
|
{
|
|
|
|
kfree(pred->field_name);
|
|
|
|
pred->field_name = NULL;
|
|
|
|
pred->str_len = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_set_pred(struct filter_pred *dest,
|
|
|
|
struct filter_pred *src,
|
|
|
|
filter_pred_fn_t fn)
|
|
|
|
{
|
|
|
|
*dest = *src;
|
|
|
|
dest->field_name = kstrdup(src->field_name, GFP_KERNEL);
|
|
|
|
if (!dest->field_name)
|
|
|
|
return -ENOMEM;
|
|
|
|
dest->fn = fn;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
static void __filter_disable_preds(struct ftrace_event_call *call)
|
2009-03-22 08:31:04 +00:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
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 08:17:50 +00:00
|
|
|
call->n_preds = 0;
|
|
|
|
|
|
|
|
for (i = 0; i < MAX_FILTER_PRED; i++)
|
|
|
|
call->preds[i]->fn = filter_pred_none;
|
|
|
|
}
|
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
void filter_disable_preds(struct ftrace_event_call *call)
|
|
|
|
{
|
|
|
|
mutex_lock(&filter_mutex);
|
|
|
|
__filter_disable_preds(call);
|
|
|
|
mutex_unlock(&filter_mutex);
|
|
|
|
}
|
|
|
|
|
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 08:17:50 +00:00
|
|
|
int init_preds(struct ftrace_event_call *call)
|
|
|
|
{
|
|
|
|
struct filter_pred *pred;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
call->n_preds = 0;
|
|
|
|
|
|
|
|
call->preds = kzalloc(MAX_FILTER_PRED * sizeof(pred), GFP_KERNEL);
|
|
|
|
if (!call->preds)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
for (i = 0; i < MAX_FILTER_PRED; i++) {
|
|
|
|
pred = kzalloc(sizeof(*pred), GFP_KERNEL);
|
|
|
|
if (!pred)
|
|
|
|
goto oom;
|
|
|
|
pred->fn = filter_pred_none;
|
|
|
|
call->preds[i] = pred;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
oom:
|
|
|
|
for (i = 0; i < MAX_FILTER_PRED; i++) {
|
|
|
|
if (call->preds[i])
|
2009-03-22 08:31:04 +00:00
|
|
|
filter_free_pred(call->preds[i]);
|
|
|
|
}
|
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 08:17:50 +00:00
|
|
|
kfree(call->preds);
|
|
|
|
call->preds = NULL;
|
|
|
|
|
|
|
|
return -ENOMEM;
|
2009-03-22 08:31:04 +00:00
|
|
|
}
|
2009-04-10 22:12:50 +00:00
|
|
|
EXPORT_SYMBOL_GPL(init_preds);
|
2009-03-22 08:31:04 +00:00
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
static void __filter_free_subsystem_preds(struct event_subsystem *system)
|
2009-03-22 08:31:17 +00:00
|
|
|
{
|
2009-04-10 17:52:20 +00:00
|
|
|
struct ftrace_event_call *call;
|
2009-03-22 08:31:17 +00:00
|
|
|
int i;
|
|
|
|
|
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 08:17:50 +00:00
|
|
|
if (system->n_preds) {
|
|
|
|
for (i = 0; i < system->n_preds; i++)
|
2009-03-22 08:31:17 +00:00
|
|
|
filter_free_pred(system->preds[i]);
|
|
|
|
kfree(system->preds);
|
|
|
|
system->preds = NULL;
|
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 08:17:50 +00:00
|
|
|
system->n_preds = 0;
|
2009-03-22 08:31:17 +00:00
|
|
|
}
|
|
|
|
|
2009-04-10 17:52:20 +00:00
|
|
|
list_for_each_entry(call, &ftrace_events, list) {
|
2009-03-31 05:48:49 +00:00
|
|
|
if (!call->define_fields)
|
2009-03-22 08:31:17 +00:00
|
|
|
continue;
|
|
|
|
|
|
|
|
if (!strcmp(call->system, system->name))
|
2009-04-17 05:27:08 +00:00
|
|
|
__filter_disable_preds(call);
|
2009-03-22 08:31:17 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
void filter_free_subsystem_preds(struct event_subsystem *system)
|
|
|
|
{
|
|
|
|
mutex_lock(&filter_mutex);
|
|
|
|
__filter_free_subsystem_preds(system);
|
|
|
|
mutex_unlock(&filter_mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int filter_add_pred_fn(struct ftrace_event_call *call,
|
|
|
|
struct filter_pred *pred,
|
|
|
|
filter_pred_fn_t fn)
|
2009-03-22 08:31:04 +00:00
|
|
|
{
|
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 08:17:50 +00:00
|
|
|
int idx, err;
|
2009-03-22 08:31:04 +00:00
|
|
|
|
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 08:17:50 +00:00
|
|
|
if (call->n_preds && !pred->compound)
|
2009-04-17 05:27:08 +00:00
|
|
|
__filter_disable_preds(call);
|
2009-03-22 08:31:04 +00:00
|
|
|
|
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 08:17:50 +00:00
|
|
|
if (call->n_preds == MAX_FILTER_PRED)
|
|
|
|
return -ENOSPC;
|
2009-03-22 08:31:04 +00:00
|
|
|
|
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 08:17:50 +00:00
|
|
|
idx = call->n_preds;
|
|
|
|
filter_clear_pred(call->preds[idx]);
|
|
|
|
err = filter_set_pred(call->preds[idx], pred, fn);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
call->n_preds++;
|
2009-03-22 08:31:04 +00:00
|
|
|
|
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 08:17:50 +00:00
|
|
|
return 0;
|
2009-03-22 08:31:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static int is_string_field(const char *type)
|
|
|
|
{
|
|
|
|
if (strchr(type, '[') && strstr(type, "char"))
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
static int __filter_add_pred(struct ftrace_event_call *call,
|
|
|
|
struct filter_pred *pred)
|
2009-03-22 08:31:04 +00:00
|
|
|
{
|
|
|
|
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 08:17:50 +00:00
|
|
|
filter_pred_fn_t fn;
|
2009-04-21 09:12:11 +00:00
|
|
|
unsigned long long val;
|
2009-03-22 08:31:04 +00:00
|
|
|
|
|
|
|
field = find_event_field(call, pred->field_name);
|
|
|
|
if (!field)
|
|
|
|
return -EINVAL;
|
|
|
|
|
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 08:17:50 +00:00
|
|
|
pred->fn = filter_pred_none;
|
2009-03-22 08:31:04 +00:00
|
|
|
pred->offset = field->offset;
|
|
|
|
|
|
|
|
if (is_string_field(field->type)) {
|
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 08:17:50 +00:00
|
|
|
fn = filter_pred_string;
|
2009-03-22 08:31:04 +00:00
|
|
|
pred->str_len = field->size;
|
2009-04-17 05:27:08 +00:00
|
|
|
return filter_add_pred_fn(call, pred, fn);
|
2009-03-24 07:14:42 +00:00
|
|
|
} else {
|
2009-04-21 09:12:11 +00:00
|
|
|
if (strict_strtoull(pred->str_val, 0, &val))
|
2009-03-24 07:14:42 +00:00
|
|
|
return -EINVAL;
|
2009-04-21 09:12:11 +00:00
|
|
|
pred->val = val;
|
2009-03-22 08:31:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
switch (field->size) {
|
|
|
|
case 8:
|
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 08:17:50 +00:00
|
|
|
fn = filter_pred_64;
|
2009-03-22 08:31:04 +00:00
|
|
|
break;
|
|
|
|
case 4:
|
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 08:17:50 +00:00
|
|
|
fn = filter_pred_32;
|
2009-03-22 08:31:04 +00:00
|
|
|
break;
|
|
|
|
case 2:
|
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 08:17:50 +00:00
|
|
|
fn = filter_pred_16;
|
2009-03-22 08:31:04 +00:00
|
|
|
break;
|
|
|
|
case 1:
|
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 08:17:50 +00:00
|
|
|
fn = filter_pred_8;
|
2009-03-22 08:31:04 +00:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
return filter_add_pred_fn(call, pred, fn);
|
|
|
|
}
|
|
|
|
|
|
|
|
int filter_add_pred(struct ftrace_event_call *call, struct filter_pred *pred)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
|
|
|
mutex_lock(&filter_mutex);
|
|
|
|
err = __filter_add_pred(call, pred);
|
|
|
|
mutex_unlock(&filter_mutex);
|
|
|
|
|
|
|
|
return err;
|
2009-03-22 08:31:17 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
int filter_add_subsystem_pred(struct event_subsystem *system,
|
|
|
|
struct filter_pred *pred)
|
|
|
|
{
|
2009-04-10 17:52:20 +00:00
|
|
|
struct ftrace_event_call *call;
|
2009-03-22 08:31:17 +00:00
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
mutex_lock(&filter_mutex);
|
|
|
|
|
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 08:17:50 +00:00
|
|
|
if (system->n_preds && !pred->compound)
|
2009-04-17 05:27:08 +00:00
|
|
|
__filter_free_subsystem_preds(system);
|
2009-03-22 08:31:17 +00:00
|
|
|
|
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 08:17:50 +00:00
|
|
|
if (!system->n_preds) {
|
2009-03-22 08:31:17 +00:00
|
|
|
system->preds = kzalloc(MAX_FILTER_PRED * sizeof(pred),
|
|
|
|
GFP_KERNEL);
|
2009-04-17 05:27:08 +00:00
|
|
|
if (!system->preds) {
|
|
|
|
mutex_unlock(&filter_mutex);
|
2009-03-22 08:31:17 +00:00
|
|
|
return -ENOMEM;
|
2009-04-17 05:27:08 +00:00
|
|
|
}
|
2009-03-22 08:31:17 +00:00
|
|
|
}
|
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
if (system->n_preds == MAX_FILTER_PRED) {
|
|
|
|
mutex_unlock(&filter_mutex);
|
2009-04-11 07:55:28 +00:00
|
|
|
return -ENOSPC;
|
2009-04-17 05:27:08 +00:00
|
|
|
}
|
2009-03-23 08:26:48 +00:00
|
|
|
|
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 08:17:50 +00:00
|
|
|
system->preds[system->n_preds] = pred;
|
2009-04-17 05:27:08 +00:00
|
|
|
system->n_preds++;
|
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 08:17:50 +00:00
|
|
|
|
2009-04-10 17:52:20 +00:00
|
|
|
list_for_each_entry(call, &ftrace_events, list) {
|
2009-03-23 08:26:48 +00:00
|
|
|
int err;
|
|
|
|
|
2009-03-31 05:48:49 +00:00
|
|
|
if (!call->define_fields)
|
2009-03-22 08:31:17 +00:00
|
|
|
continue;
|
|
|
|
|
2009-03-23 08:26:48 +00:00
|
|
|
if (strcmp(call->system, system->name))
|
|
|
|
continue;
|
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
err = __filter_add_pred(call, 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 08:17:50 +00:00
|
|
|
if (err == -ENOMEM) {
|
|
|
|
system->preds[system->n_preds] = NULL;
|
2009-04-17 05:27:08 +00:00
|
|
|
system->n_preds--;
|
|
|
|
mutex_unlock(&filter_mutex);
|
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 08:17:50 +00:00
|
|
|
return err;
|
|
|
|
}
|
2009-03-22 08:31:17 +00:00
|
|
|
}
|
|
|
|
|
2009-04-17 05:27:08 +00:00
|
|
|
mutex_unlock(&filter_mutex);
|
2009-03-23 08:26:48 +00:00
|
|
|
|
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 08:17:50 +00:00
|
|
|
return 0;
|
2009-03-22 08:31:17 +00:00
|
|
|
}
|
|
|
|
|
2009-04-21 09:12:11 +00:00
|
|
|
/*
|
|
|
|
* The filter format can be
|
|
|
|
* - 0, which means remove all filter preds
|
|
|
|
* - [||/&&] <field> ==/!= <val>
|
|
|
|
*/
|
2009-03-22 08:31:04 +00:00
|
|
|
int filter_parse(char **pbuf, struct filter_pred *pred)
|
|
|
|
{
|
2009-04-21 09:12:11 +00:00
|
|
|
char *tok, *val_str = NULL;
|
2009-03-22 08:31:04 +00:00
|
|
|
int tok_n = 0;
|
|
|
|
|
|
|
|
while ((tok = strsep(pbuf, " \n"))) {
|
|
|
|
if (tok_n == 0) {
|
|
|
|
if (!strcmp(tok, "0")) {
|
|
|
|
pred->clear = 1;
|
|
|
|
return 0;
|
|
|
|
} else if (!strcmp(tok, "&&")) {
|
|
|
|
pred->or = 0;
|
|
|
|
pred->compound = 1;
|
|
|
|
} else if (!strcmp(tok, "||")) {
|
|
|
|
pred->or = 1;
|
|
|
|
pred->compound = 1;
|
|
|
|
} else
|
|
|
|
pred->field_name = tok;
|
|
|
|
tok_n = 1;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if (tok_n == 1) {
|
|
|
|
if (!pred->field_name)
|
|
|
|
pred->field_name = tok;
|
|
|
|
else if (!strcmp(tok, "!="))
|
|
|
|
pred->not = 1;
|
|
|
|
else if (!strcmp(tok, "=="))
|
|
|
|
pred->not = 0;
|
|
|
|
else {
|
|
|
|
pred->field_name = NULL;
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
tok_n = 2;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if (tok_n == 2) {
|
|
|
|
if (pred->compound) {
|
|
|
|
if (!strcmp(tok, "!="))
|
|
|
|
pred->not = 1;
|
|
|
|
else if (!strcmp(tok, "=="))
|
|
|
|
pred->not = 0;
|
|
|
|
else {
|
|
|
|
pred->field_name = NULL;
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
val_str = tok;
|
|
|
|
break; /* done */
|
|
|
|
}
|
|
|
|
tok_n = 3;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if (tok_n == 3) {
|
|
|
|
val_str = tok;
|
|
|
|
break; /* done */
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
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 08:17:50 +00:00
|
|
|
if (!val_str || !strlen(val_str)
|
|
|
|
|| strlen(val_str) >= MAX_FILTER_STR_VAL) {
|
2009-04-11 07:52:35 +00:00
|
|
|
pred->field_name = NULL;
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2009-04-21 09:12:11 +00:00
|
|
|
strcpy(pred->str_val, val_str);
|
|
|
|
pred->str_len = strlen(val_str);
|
|
|
|
|
2009-03-22 08:31:04 +00:00
|
|
|
pred->field_name = kstrdup(pred->field_name, GFP_KERNEL);
|
|
|
|
if (!pred->field_name)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
|