Add dtnfsclient, a first cut at an NFSv2/v3 client reuest DTrace

provider.  The NFS client exposes 'start' and 'done' probes for NFSv2
and NFSv3 RPCs when using the new RPC implementation, passing in the
vnode, mbuf chain, credential, and NFSv2 or NFSv3 procedure number.
For 'done' probes, the error number is also available.

Probes are named in the following way:

  ...
  nfsclient:nfs2:write:start
  nfsclient:nfs2:write:done
  ...
  nfsclient:nfs3:access:start
  nfsclient:nfs3:access:done
  ...

Access to the unmarshalled arguments is not easily available at this
point in the stack, but the passed probe arguments are sufficient to
to a lot of interesting things in practice.  Technically, these probes
may cover multiple RPC retransmits, and even transactions if the
transaction ID change as a result of authentication failure or a
jukebox error from the server, but usefully capture the intent of a
single NFS request, such as access, getattr, write, etc.

Typical use might involve profiling RPC latency by system call, number
of RPCs, how often a getattr leads to a call to access, when failed
access control checks occur, etc.  More detailed RPC information might
best be provided by adding a krpc provider.  It would also be useful
to add NFS client probes for events such as the access cache or
attribute cache satisfying requests without an RPC.

Sponsored by:	Google, Inc.
MFC after:	1 month
This commit is contained in:
Robert Watson 2009-03-22 22:07:52 +00:00
parent d956d1eed5
commit 47294818f9
Notes: svn2git 2020-12-20 02:59:44 +00:00
svn path=/head/; revision=190293
6 changed files with 445 additions and 0 deletions

View File

@ -0,0 +1,343 @@
/*-
* Copyright (c) 2009 Robert N. M. Watson
* All rights reserved.
*
* This software was developed at the University of Cambridge Computer
* Laboratory with support from a grant from Google, Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/conf.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/dtrace.h>
#include <sys/dtrace_bsd.h>
#include <nfs/nfsproto.h>
/*
* dtnfsclient is a DTrace provider that tracks the intent to perform RPCs
* in the NFS client. This is not quite the same as RPCs, because NFS may
* issue multiple RPC transactions in the event that authentication fails,
* there's a jukebox error, etc. However, it cleanly represents the logical
* layer between RPC transmission and vnode/vfs operations, providing access
* to state linking the two.
*/
static int dtnfsclient_unload(void);
static void dtnfsclient_getargdesc(void *, dtrace_id_t, void *,
dtrace_argdesc_t *);
static void dtnfsclient_provide(void *, dtrace_probedesc_t *);
static void dtnfsclient_destroy(void *, dtrace_id_t, void *);
static void dtnfsclient_enable(void *, dtrace_id_t, void *);
static void dtnfsclient_disable(void *, dtrace_id_t, void *);
static void dtnfsclient_load(void *);
static dtrace_pattr_t dtnfsclient_attr = {
{ DTRACE_STABILITY_STABLE, DTRACE_STABILITY_STABLE, DTRACE_CLASS_COMMON },
{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
{ DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
{ DTRACE_STABILITY_STABLE, DTRACE_STABILITY_STABLE, DTRACE_CLASS_COMMON },
{ DTRACE_STABILITY_STABLE, DTRACE_STABILITY_STABLE, DTRACE_CLASS_COMMON },
};
struct dtnfsclient_rpc {
char *nr_v3_name;
char *nr_v2_name; /* Or NULL if none. */
/*
* IDs for the start and done cases, for both NFSv2 and NFSv3.
*/
uint32_t nr_v2_id_start, nr_v2_id_done;
uint32_t nr_v3_id_start, nr_v3_id_done;
};
/*
* This table is indexed by NFSv3 procedure number, but also used for NFSv2
* procedure names.
*/
static struct dtnfsclient_rpc dtnfsclient_rpcs[NFS_NPROCS] = {
{ "null", "null" },
{ "getattr", "getattr" },
{ "setattr", "setattr" },
{ "lookup", "lookup" },
{ "access" },
{ "readlink", "readlink" },
{ "read", "read" },
{ "write", "write" },
{ "create", "create" },
{ "mkdir", "mkdir" },
{ "symlink", "symlink" },
{ "mknod" },
{ "remove", "remove" },
{ "rmdir", "rmdir" },
{ "rename", "rename" },
{ "link", "link" },
{ "readdir", "readdir" },
{ "readdirplus" },
{ "fsstat", "statfs" },
{ "fsinfo" },
{ "pathconf" },
{ "commit" },
{ "noop" },
};
static char *dtnfsclient_v2modulename = "nfs2";
static char *dtnfsclient_v3modulename = "nfs3";
static char *dtnfsclient_start = "start";
static char *dtnfsclient_done = "done";
static dtrace_pops_t dtnfsclient_pops = {
dtnfsclient_provide,
NULL,
dtnfsclient_enable,
dtnfsclient_disable,
NULL,
NULL,
dtnfsclient_getargdesc,
NULL,
NULL,
dtnfsclient_destroy
};
static dtrace_provider_id_t dtnfsclient_id;
/*
* When tracing on a procedure is enabled, the DTrace ID for the event is
* stored in one of these two NFS client-allocated arrays; 0 indicates that
* the event is not being traced so probes should not be called.
*
* For simplicity, we allocate both v2 and v3 arrays as NFS_NPROCS, and the
* v2 array is simply sparse.
*/
extern uint32_t nfsclient_nfs2_start_probes[NFS_NPROCS];
extern uint32_t nfsclient_nfs2_done_probes[NFS_NPROCS];
extern uint32_t nfsclient_nfs3_start_probes[NFS_NPROCS];
extern uint32_t nfsclient_nfs3_done_probes[NFS_NPROCS];
/*
* Look up a DTrace probe ID to see if it's associated with a "done" event --
* if so, we will return a fourth argument type of "int".
*/
static int
dtnfs23_isdoneprobe(dtrace_id_t id)
{
int i;
for (i = 0; i < NFS_NPROCS; i++) {
if (dtnfsclient_rpcs[i].nr_v3_id_done == id ||
dtnfsclient_rpcs[i].nr_v2_id_done == id)
return (1);
}
return (0);
}
static void
dtnfsclient_getargdesc(void *arg, dtrace_id_t id, void *parg,
dtrace_argdesc_t *desc)
{
const char *p = NULL;
switch (desc->dtargd_ndx) {
case 0:
p = "struct vnode *";
break;
case 1:
p = "struct mbuf *";
break;
case 2:
p = "struct ucred *";
break;
case 3:
p = "int";
break;
case 4:
if (dtnfs23_isdoneprobe(id)) {
p = "int";
break;
}
/* FALLSTHROUGH */
default:
desc->dtargd_ndx = DTRACE_ARGNONE;
break;
}
if (p != NULL)
strlcpy(desc->dtargd_native, p, sizeof(desc->dtargd_native));
}
static void
dtnfsclient_provide(void *arg, dtrace_probedesc_t *desc)
{
int i;
if (desc != NULL)
return;
/*
* First, register NFSv2 RPC procedures; note sparseness check for
* each slot in the NFSv3 procnum-indexed array.
*/
for (i = 0; i < NFS_NPROCS; i++) {
if (dtnfsclient_rpcs[i].nr_v2_name != NULL &&
dtrace_probe_lookup(dtnfsclient_id,
dtnfsclient_v2modulename, dtnfsclient_rpcs[i].nr_v2_name,
dtnfsclient_start) == 0) {
dtnfsclient_rpcs[i].nr_v2_id_start =
dtrace_probe_create(dtnfsclient_id,
dtnfsclient_v2modulename,
dtnfsclient_rpcs[i].nr_v2_name,
dtnfsclient_start, 0,
&nfsclient_nfs2_start_probes[i]);
}
if (dtnfsclient_rpcs[i].nr_v2_name != NULL &&
dtrace_probe_lookup(dtnfsclient_id,
dtnfsclient_v2modulename, dtnfsclient_rpcs[i].nr_v2_name,
dtnfsclient_done) == 0) {
dtnfsclient_rpcs[i].nr_v2_id_done =
dtrace_probe_create(dtnfsclient_id,
dtnfsclient_v2modulename,
dtnfsclient_rpcs[i].nr_v2_name,
dtnfsclient_done, 0,
&nfsclient_nfs2_done_probes[i]);
}
}
/*
* Now, register NFSv3 RPC procedures.
*/
for (i = 0; i < NFS_NPROCS; i++) {
if (dtrace_probe_lookup(dtnfsclient_id,
dtnfsclient_v3modulename, dtnfsclient_rpcs[i].nr_v3_name,
dtnfsclient_start) == 0) {
dtnfsclient_rpcs[i].nr_v3_id_start =
dtrace_probe_create(dtnfsclient_id,
dtnfsclient_v3modulename,
dtnfsclient_rpcs[i].nr_v3_name,
dtnfsclient_start, 0,
&nfsclient_nfs3_start_probes[i]);
}
if (dtrace_probe_lookup(dtnfsclient_id,
dtnfsclient_v3modulename, dtnfsclient_rpcs[i].nr_v3_name,
dtnfsclient_done) == 0) {
dtnfsclient_rpcs[i].nr_v3_id_done =
dtrace_probe_create(dtnfsclient_id,
dtnfsclient_v3modulename,
dtnfsclient_rpcs[i].nr_v3_name,
dtnfsclient_done, 0,
&nfsclient_nfs3_done_probes[i]);
}
}
}
static void
dtnfsclient_destroy(void *arg, dtrace_id_t id, void *parg)
{
}
static void
dtnfsclient_enable(void *arg, dtrace_id_t id, void *parg)
{
uint32_t *p = parg;
*p = id;
}
static void
dtnfsclient_disable(void *arg, dtrace_id_t id, void *parg)
{
uint32_t *p = parg;
*p = 0;
}
static void
dtnfsclient_load(void *dummy)
{
if (dtrace_register("nfsclient", &dtnfsclient_attr,
DTRACE_PRIV_USER, NULL, &dtnfsclient_pops, NULL,
&dtnfsclient_id) != 0)
return;
dtrace_nfsclient_nfs23_start_probe =
(dtrace_nfsclient_nfs23_start_probe_func_t)dtrace_probe;
dtrace_nfsclient_nfs23_done_probe =
(dtrace_nfsclient_nfs23_done_probe_func_t)dtrace_probe;
}
static int
dtnfsclient_unload()
{
int error = 0;
dtrace_nfsclient_nfs23_start_probe = NULL;
dtrace_nfsclient_nfs23_done_probe = NULL;
if ((error = dtrace_unregister(dtnfsclient_id)) != 0)
return (error);
return (error);
}
static int
dtnfsclient_modevent(module_t mod __unused, int type, void *data __unused)
{
int error = 0;
switch (type) {
case MOD_LOAD:
break;
case MOD_UNLOAD:
break;
case MOD_SHUTDOWN:
break;
default:
error = EOPNOTSUPP;
break;
}
return (error);
}
SYSINIT(dtnfsclient_load, SI_SUB_DTRACE_PROVIDER, SI_ORDER_ANY,
dtnfsclient_load, NULL);
SYSUNINIT(dtnfsclient_unload, SI_SUB_DTRACE_PROVIDER, SI_ORDER_ANY,
dtnfsclient_unload, NULL);
DEV_MODULE(dtnfsclient, dtnfsclient_modevent, NULL);
MODULE_VERSION(dtnfsclient, 1);
MODULE_DEPEND(dtnfsclient, dtrace, 1, 1, 1);
MODULE_DEPEND(dtnfsclient, opensolaris, 1, 1, 1);

View File

@ -4,6 +4,7 @@
.include "Makefile.inc"
SUBDIR= dtmalloc \
dtnfsclient \
dtrace \
dtraceall \
dtrace_test \

View File

@ -0,0 +1,13 @@
# $FreeBSD$
.PATH: ${.CURDIR}/../../../cddl/dev/dtnfsclient
KMOD= dtnfsclient
SRCS= dtnfsclient.c
SRCS+= vnode_if.h
CFLAGS+= -I${.CURDIR}/../../../cddl/compat/opensolaris \
-I${.CURDIR}/../../../cddl/contrib/opensolaris/uts/common \
-I${.CURDIR}/../../..
.include <bsd.kmod.mk>

View File

@ -65,6 +65,7 @@ MODULE_DEPEND(dtraceall, cyclic, 1, 1, 1);
MODULE_DEPEND(dtraceall, opensolaris, 1, 1, 1);
MODULE_DEPEND(dtraceall, dtrace, 1, 1, 1);
MODULE_DEPEND(dtraceall, dtmalloc, 1, 1, 1);
MODULE_DEPEND(dtraceall, dtnfsclient, 1, 1, 1);
#if defined(__amd64__) || defined(__i386__)
MODULE_DEPEND(dtraceall, fbt, 1, 1, 1);
#endif

View File

@ -40,6 +40,7 @@ __FBSDID("$FreeBSD$");
*/
#include "opt_inet6.h"
#include "opt_kdtrace.h"
#include "opt_kgssapi.h"
#include <sys/param.h>
@ -73,6 +74,25 @@ __FBSDID("$FreeBSD$");
#ifndef NFS_LEGACYRPC
#ifdef KDTRACE_HOOKS
#include <sys/dtrace_bsd.h>
dtrace_nfsclient_nfs23_start_probe_func_t
dtrace_nfsclient_nfs23_start_probe;
dtrace_nfsclient_nfs23_done_probe_func_t
dtrace_nfsclient_nfs23_done_probe;
/*
* Registered probes by RPC type.
*/
uint32_t nfsclient_nfs2_start_probes[NFS_NPROCS];
uint32_t nfsclient_nfs2_done_probes[NFS_NPROCS];
uint32_t nfsclient_nfs3_start_probes[NFS_NPROCS];
uint32_t nfsclient_nfs3_done_probes[NFS_NPROCS];
#endif
static int nfs_realign_test;
static int nfs_realign_count;
static int nfs_bufpackets = 4;
@ -468,6 +488,25 @@ nfs_request(struct vnode *vp, struct mbuf *mreq, int procnum,
ext.rc_timers = NULL;
}
#ifdef KDTRACE_HOOKS
if (dtrace_nfsclient_nfs23_start_probe != NULL) {
uint32_t probe_id;
int probe_procnum;
if (nmp->nm_flag & NFSMNT_NFSV3) {
probe_id = nfsclient_nfs3_start_probes[procnum];
probe_procnum = procnum;
} else {
probe_id = nfsclient_nfs2_start_probes[procnum];
probe_procnum = (nmp->nm_flag & NFSMNT_NFSV3) ?
procnum : nfsv2_procid[procnum];
}
if (probe_id != 0)
(dtrace_nfsclient_nfs23_start_probe)(probe_id, vp,
mreq, cred, probe_procnum);
}
#endif
nfsstats.rpcrequests++;
tryagain:
timo.tv_sec = nmp->nm_timeo / NFS_HZ;
@ -535,6 +574,24 @@ nfs_request(struct vnode *vp, struct mbuf *mreq, int procnum,
goto nfsmout;
}
#ifdef KDTRACE_HOOKS
if (dtrace_nfsclient_nfs23_done_probe != NULL) {
uint32_t probe_id;
int probe_procnum;
if (nmp->nm_flag & NFSMNT_NFSV3) {
probe_id = nfsclient_nfs3_done_probes[procnum];
probe_procnum = procnum;
} else {
probe_id = nfsclient_nfs2_done_probes[procnum];
probe_procnum = (nmp->nm_flag & NFSMNT_NFSV3) ?
procnum : nfsv2_procid[procnum];
}
if (probe_id != 0)
(dtrace_nfsclient_nfs23_done_probe)(probe_id, vp,
mreq, cred, probe_procnum, 0);
}
#endif
m_freem(mreq);
*mrp = mrep;
*mdp = md;
@ -543,6 +600,24 @@ nfs_request(struct vnode *vp, struct mbuf *mreq, int procnum,
return (0);
nfsmout:
#ifdef KDTRACE_HOOKS
if (dtrace_nfsclient_nfs23_done_probe != NULL) {
uint32_t probe_id;
int probe_procnum;
if (nmp->nm_flag & NFSMNT_NFSV3) {
probe_id = nfsclient_nfs3_done_probes[procnum];
probe_procnum = procnum;
} else {
probe_id = nfsclient_nfs2_done_probes[procnum];
probe_procnum = (nmp->nm_flag & NFSMNT_NFSV3) ?
procnum : nfsv2_procid[procnum];
}
if (probe_id != 0)
(dtrace_nfsclient_nfs23_done_probe)(probe_id, vp,
mreq, cred, probe_procnum, error);
}
#endif
m_freem(mreq);
if (auth)
AUTH_DESTROY(auth);

View File

@ -32,6 +32,7 @@
#define _SYS_DTRACE_BSD_H
/* Forward definitions: */
struct mbuf;
struct trapframe;
struct thread;
@ -93,6 +94,17 @@ typedef void (*dtrace_malloc_probe_func_t)(u_int32_t, uintptr_t arg0,
extern dtrace_malloc_probe_func_t dtrace_malloc_probe;
/* The dtnfsclient provider hooks into the NFS[23] client. */
typedef void (*dtrace_nfsclient_nfs23_start_probe_func_t)(u_int32_t,
struct vnode *, struct mbuf *, struct ucred *, int);
typedef void (*dtrace_nfsclient_nfs23_done_probe_func_t)(u_int32_t,
struct vnode *, struct mbuf *, struct ucred *, int, int);
extern dtrace_nfsclient_nfs23_start_probe_func_t
dtrace_nfsclient_nfs23_start_probe;
extern dtrace_nfsclient_nfs23_done_probe_func_t
dtrace_nfsclient_nfs23_done_probe;
/*
* Functions which allow the dtrace module to check that the kernel
* hooks have been compiled with sufficient space for it's private