freebsd-skq/sys/cddl/dev/sdt/sdt.c

404 lines
11 KiB
C
Raw Normal View History

/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*
* Portions Copyright 2006-2008 John Birrell jb@freebsd.org
*
* $FreeBSD$
*
*/
/*
* This file contains a reimplementation of the statically-defined tracing (SDT)
* framework for DTrace. Probes and SDT providers are defined using the macros
* in sys/sdt.h, which append all the needed structures to linker sets. When
* this module is loaded, it iterates over all of the loaded modules and
* registers probes and providers with the DTrace framework based on the
* contents of these linker sets.
*
* A list of SDT providers is maintained here since a provider may span multiple
* modules. When a kernel module is unloaded, a provider defined in that module
* is unregistered only if no other modules refer to it. The DTrace framework is
* responsible for destroying individual probes when a kernel module is
* unloaded; in particular, probes may not span multiple kernel modules.
*/
#include <sys/cdefs.h>
#include <sys/param.h>
#include <sys/systm.h>
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
#include <sys/conf.h>
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
#include <sys/eventhandler.h>
#include <sys/kernel.h>
#include <sys/limits.h>
#include <sys/linker.h>
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
#include <sys/linker_set.h>
#include <sys/lock.h>
#include <sys/lockstat.h>
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/mutex.h>
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
#include <sys/queue.h>
#include <sys/sdt.h>
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
#include <sys/dtrace.h>
#include <sys/dtrace_bsd.h>
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
/* DTrace methods. */
static void sdt_getargdesc(void *, dtrace_id_t, void *, dtrace_argdesc_t *);
static void sdt_provide_probes(void *, dtrace_probedesc_t *);
static void sdt_destroy(void *, dtrace_id_t, void *);
static void sdt_enable(void *, dtrace_id_t, void *);
static void sdt_disable(void *, dtrace_id_t, void *);
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
static void sdt_load(void);
static int sdt_unload(void);
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
static void sdt_create_provider(struct sdt_provider *);
static void sdt_create_probe(struct sdt_probe *);
static void sdt_kld_load(void *, struct linker_file *);
static void sdt_kld_unload_try(void *, struct linker_file *, int *);
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
static MALLOC_DEFINE(M_SDT, "SDT", "DTrace SDT providers");
static dtrace_pattr_t sdt_attr = {
{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_ISA },
{ DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_ISA },
};
static dtrace_pops_t sdt_pops = {
sdt_provide_probes,
NULL,
sdt_enable,
sdt_disable,
NULL,
NULL,
sdt_getargdesc,
NULL,
NULL,
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
sdt_destroy,
};
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
static TAILQ_HEAD(, sdt_provider) sdt_prov_list;
static eventhandler_tag sdt_kld_load_tag;
static eventhandler_tag sdt_kld_unload_try_tag;
static void
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
sdt_create_provider(struct sdt_provider *prov)
{
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
struct sdt_provider *curr, *newprov;
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
TAILQ_FOREACH(curr, &sdt_prov_list, prov_entry)
if (strcmp(prov->name, curr->name) == 0) {
/* The provider has already been defined. */
curr->sdt_refs++;
return;
}
/*
* Make a copy of prov so that we don't lose fields if its module is
* unloaded but the provider isn't destroyed. This could happen with
* a provider that spans multiple modules.
*/
newprov = malloc(sizeof(*newprov), M_SDT, M_WAITOK | M_ZERO);
newprov->name = strdup(prov->name, M_SDT);
prov->sdt_refs = newprov->sdt_refs = 1;
TAILQ_INSERT_TAIL(&sdt_prov_list, newprov, prov_entry);
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
(void)dtrace_register(newprov->name, &sdt_attr, DTRACE_PRIV_USER, NULL,
&sdt_pops, NULL, (dtrace_provider_id_t *)&newprov->id);
prov->id = newprov->id;
}
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
static void
sdt_create_probe(struct sdt_probe *probe)
{
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
struct sdt_provider *prov;
char mod[DTRACE_MODNAMELEN];
char func[DTRACE_FUNCNAMELEN];
char name[DTRACE_NAMELEN];
const char *from;
char *to;
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
size_t len;
if (probe->version != (int)sizeof(*probe)) {
printf("ignoring probe %p, version %u expected %u\n",
probe, probe->version, (int)sizeof(*probe));
return;
}
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
TAILQ_FOREACH(prov, &sdt_prov_list, prov_entry)
if (strcmp(prov->name, probe->prov->name) == 0)
break;
KASSERT(prov != NULL, ("probe defined without a provider"));
/* If no module name was specified, use the module filename. */
if (*probe->mod == 0) {
len = strlcpy(mod, probe->sdtp_lf->filename, sizeof(mod));
if (len > 3 && strcmp(mod + len - 3, ".ko") == 0)
mod[len - 3] = '\0';
} else
strlcpy(mod, probe->mod, sizeof(mod));
/*
* Unfortunately this is necessary because the Solaris DTrace
* code mixes consts and non-consts with casts to override
* the incompatibilies. On FreeBSD, we use strict warnings
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
* in the C compiler, so we have to respect const vs non-const.
*/
strlcpy(func, probe->func, sizeof(func));
if (func[0] == '\0')
strcpy(func, "none");
from = probe->name;
to = name;
for (len = 0; len < (sizeof(name) - 1) && *from != '\0';
len++, from++, to++) {
if (from[0] == '_' && from[1] == '_') {
*to = '-';
from++;
} else
*to = *from;
}
*to = '\0';
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
if (dtrace_probe_lookup(prov->id, mod, func, name) != DTRACE_IDNONE)
return;
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
(void)dtrace_probe_create(prov->id, mod, func, name, 1, probe);
}
/*
* Probes are created through the SDT module load/unload hook, so this function
* has nothing to do. It only exists because the DTrace provider framework
* requires one of provide_probes and provide_module to be defined.
*/
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
static void
sdt_provide_probes(void *arg, dtrace_probedesc_t *desc)
{
}
static void
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
sdt_enable(void *arg __unused, dtrace_id_t id, void *parg)
{
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
struct sdt_probe *probe = parg;
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
probe->id = id;
probe->sdtp_lf->nenabled++;
if (strcmp(probe->prov->name, "lockstat") == 0)
lockstat_enabled++;
}
static void
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
sdt_disable(void *arg __unused, dtrace_id_t id, void *parg)
{
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
struct sdt_probe *probe = parg;
KASSERT(probe->sdtp_lf->nenabled > 0, ("no probes enabled"));
if (strcmp(probe->prov->name, "lockstat") == 0)
lockstat_enabled--;
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
probe->id = 0;
probe->sdtp_lf->nenabled--;
}
static void
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
sdt_getargdesc(void *arg, dtrace_id_t id, void *parg, dtrace_argdesc_t *desc)
{
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
struct sdt_argtype *argtype;
struct sdt_probe *probe = parg;
if (desc->dtargd_ndx >= probe->n_args) {
desc->dtargd_ndx = DTRACE_ARGNONE;
return;
}
TAILQ_FOREACH(argtype, &probe->argtype_list, argtype_entry) {
if (desc->dtargd_ndx == argtype->ndx) {
desc->dtargd_mapping = desc->dtargd_ndx;
if (argtype->type == NULL) {
desc->dtargd_native[0] = '\0';
desc->dtargd_xlate[0] = '\0';
continue;
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
}
strlcpy(desc->dtargd_native, argtype->type,
sizeof(desc->dtargd_native));
if (argtype->xtype != NULL)
strlcpy(desc->dtargd_xlate, argtype->xtype,
sizeof(desc->dtargd_xlate));
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
}
}
}
static void
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
sdt_destroy(void *arg, dtrace_id_t id, void *parg)
{
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
}
/*
* Called from the kernel linker when a module is loaded, before
* dtrace_module_loaded() is called. This is done so that it's possible to
* register new providers when modules are loaded. The DTrace framework
* explicitly disallows calling into the framework from the provide_module
* provider method, so we cannot do this there.
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
*/
static void
sdt_kld_load(void *arg __unused, struct linker_file *lf)
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
{
struct sdt_provider **prov, **begin, **end;
struct sdt_probe **probe, **p_begin, **p_end;
struct sdt_argtype **argtype, **a_begin, **a_end;
if (linker_file_lookup_set(lf, "sdt_providers_set", &begin, &end,
NULL) == 0) {
for (prov = begin; prov < end; prov++)
sdt_create_provider(*prov);
}
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
if (linker_file_lookup_set(lf, "sdt_probes_set", &p_begin, &p_end,
NULL) == 0) {
for (probe = p_begin; probe < p_end; probe++) {
(*probe)->sdtp_lf = lf;
sdt_create_probe(*probe);
TAILQ_INIT(&(*probe)->argtype_list);
}
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
}
if (linker_file_lookup_set(lf, "sdt_argtypes_set", &a_begin, &a_end,
NULL) == 0) {
for (argtype = a_begin; argtype < a_end; argtype++) {
(*argtype)->probe->n_args++;
TAILQ_INSERT_TAIL(&(*argtype)->probe->argtype_list,
*argtype, argtype_entry);
}
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
}
}
static void
sdt_kld_unload_try(void *arg __unused, struct linker_file *lf, int *error)
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
{
struct sdt_provider *prov, **curr, **begin, **end, *tmp;
if (*error != 0)
/* We already have an error, so don't do anything. */
return;
else if (linker_file_lookup_set(lf, "sdt_providers_set", &begin, &end,
NULL))
/* No DTrace providers are declared in this file. */
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
return;
/*
* Go through all the providers declared in this linker file and
* unregister any that aren't declared in another loaded file.
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
*/
for (curr = begin; curr < end; curr++) {
TAILQ_FOREACH_SAFE(prov, &sdt_prov_list, prov_entry, tmp) {
if (strcmp(prov->name, (*curr)->name) != 0)
continue;
if (prov->sdt_refs == 1) {
if (dtrace_unregister(prov->id) != 0) {
*error = 1;
return;
}
TAILQ_REMOVE(&sdt_prov_list, prov, prov_entry);
free(prov->name, M_SDT);
free(prov, M_SDT);
} else
prov->sdt_refs--;
break;
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
}
}
}
static int
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
sdt_linker_file_cb(linker_file_t lf, void *arg __unused)
{
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
sdt_kld_load(NULL, lf);
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
return (0);
}
static void
sdt_load()
{
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
TAILQ_INIT(&sdt_prov_list);
sdt_probe_func = dtrace_probe;
sdt_kld_load_tag = EVENTHANDLER_REGISTER(kld_load, sdt_kld_load, NULL,
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
EVENTHANDLER_PRI_ANY);
sdt_kld_unload_try_tag = EVENTHANDLER_REGISTER(kld_unload_try,
sdt_kld_unload_try, NULL, EVENTHANDLER_PRI_ANY);
/* Pick up probes from the kernel and already-loaded linker files. */
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
linker_file_foreach(sdt_linker_file_cb, NULL);
}
static int
sdt_unload()
{
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
struct sdt_provider *prov, *tmp;
int ret;
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
EVENTHANDLER_DEREGISTER(kld_load, sdt_kld_load_tag);
EVENTHANDLER_DEREGISTER(kld_unload_try, sdt_kld_unload_try_tag);
sdt_probe_func = sdt_probe_stub;
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
TAILQ_FOREACH_SAFE(prov, &sdt_prov_list, prov_entry, tmp) {
ret = dtrace_unregister(prov->id);
if (ret != 0)
return (ret);
FreeBSD's DTrace implementation has a few problems with respect to handling probes declared in a kernel module when that module is unloaded. In particular, * Unloading a module with active SDT probes will cause a panic. [1] * A module's (FBT/SDT) probes aren't destroyed when the module is unloaded; trying to use them after the fact will generally cause a panic. This change fixes both problems by porting the DTrace module load/unload handlers from illumos and registering them with the corresponding EVENTHANDLER(9) handlers. This allows the DTrace framework to destroy all probes defined in a module when that module is unloaded, and to prevent a module unload from proceeding if some of its probes are active. The latter problem has already been fixed for FBT probes by checking lf->nenabled in kern_kldunload(), but moving the check into the DTrace framework generalizes it to all kernel providers and also fixes a race in the current implementation (since a probe may be activated between the check and the call to linker_file_unload()). Additionally, the SDT implementation has been reworked to define SDT providers/probes/argtypes in linker sets rather than using SYSINIT/SYSUNINIT to create and destroy SDT probes when a module is loaded or unloaded. This simplifies things quite a bit since it means that pretty much all of the SDT code can live in sdt.ko, and since it becomes easier to integrate SDT with the DTrace framework. Furthermore, this allows FreeBSD to be quite flexible in that SDT providers spanning multiple modules can be created on the fly when a module is loaded; at the moment it looks like illumos' SDT implementation requires all SDT probes to be statically defined in a single kernel table. PR: 166927, 166926, 166928 Reported by: davide [1] Reviewed by: avg, trociny (earlier version) MFC after: 1 month
2013-08-13 03:10:39 +00:00
TAILQ_REMOVE(&sdt_prov_list, prov, prov_entry);
free(prov->name, M_SDT);
free(prov, M_SDT);
}
return (0);
}
static int
sdt_modevent(module_t mod __unused, int type, void *data __unused)
{
switch (type) {
case MOD_LOAD:
case MOD_UNLOAD:
case MOD_SHUTDOWN:
return (0);
default:
return (EOPNOTSUPP);
}
}
SYSINIT(sdt_load, SI_SUB_DTRACE_PROVIDER, SI_ORDER_ANY, sdt_load, NULL);
SYSUNINIT(sdt_unload, SI_SUB_DTRACE_PROVIDER, SI_ORDER_ANY, sdt_unload, NULL);
DEV_MODULE(sdt, sdt_modevent, NULL);
MODULE_VERSION(sdt, 1);
MODULE_DEPEND(sdt, dtrace, 1, 1, 1);