freebsd-skq/cddl/usr.sbin/zfsd/zfsd_event.cc
Alan Somers 7a0c41d5d7 zfsd(8), the ZFS fault management daemon
Add zfsd, which deals with hard drive faults in ZFS pools. It manages
hotspares and replements in drive slots that publish physical paths.

cddl/usr.sbin/zfsd
	Add zfsd(8) and its unit tests

cddl/usr.sbin/Makefile
	Add zfsd to the build

lib/libdevdctl
	A C++ library that helps devd clients process events

lib/Makefile
share/mk/bsd.libnames.mk
share/mk/src.libnames.mk
	Add libdevdctl to the build. It's a private library, unusable by
	out-of-tree software.

etc/defaults/rc.conf
	By default, set zfsd_enable to NO

etc/mtree/BSD.include.dist
	Add a directory for libdevdctl's include files

etc/mtree/BSD.tests.dist
	Add a directory for zfsd's unit tests

etc/mtree/BSD.var.dist
	Add /var/db/zfsd/cases, where zfsd stores case files while it's shut
	down.

etc/rc.d/Makefile
etc/rc.d/zfsd
	Add zfsd's rc script

sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev.c
	Fix the resource.fs.zfs.statechange message. It had a number of
	problems:

	It was only being emitted on a transition to the HEALTHY state.
	That made it impossible for zfsd to take actions based on drives
	getting sicker.

	It compared the new state to vdev_prevstate, which is the state that
	the vdev had the last time it was opened.  That doesn't make sense,
	because a vdev can change state multiple times without being
	reopened.

	vdev_set_state contains logic that will change the device's new
	state based on various conditions.  However, the statechange event
	was being posted _before_ that logic took effect.  Now it's being
	posted after.

Submitted by:	gibbs, asomers, mav, allanjude
Reviewed by:	mav, delphij
Relnotes:	yes
Sponsored by:	Spectra Logic Corp, iX Systems
Differential Revision:	https://reviews.freebsd.org/D6564
2016-05-28 17:43:40 +00:00

536 lines
14 KiB
C++

/*-
* Copyright (c) 2011, 2012, 2013, 2014, 2016 Spectra Logic Corporation
* All rights reserved.
*
* 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,
* without modification.
* 2. Redistributions in binary form must reproduce at minimum a disclaimer
* substantially similar to the "NO WARRANTY" disclaimer below
* ("Disclaimer") and any redistribution must be conditioned upon
* including a substantially similar Disclaimer requirement for further
* binary redistribution.
*
* NO WARRANTY
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* HOLDERS OR CONTRIBUTORS BE LIABLE FOR 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 DAMAGES.
*
* Authors: Justin T. Gibbs (Spectra Logic Corporation)
*/
/**
* \file zfsd_event.cc
*/
#include <sys/cdefs.h>
#include <sys/time.h>
#include <sys/fs/zfs.h>
#include <syslog.h>
#include <libzfs.h>
/*
* Undefine flush, defined by cpufunc.h on sparc64, because it conflicts with
* C++ flush methods
*/
#undef flush
#include <list>
#include <map>
#include <sstream>
#include <string>
#include <devdctl/guid.h>
#include <devdctl/event.h>
#include <devdctl/event_factory.h>
#include <devdctl/exception.h>
#include <devdctl/consumer.h>
#include "callout.h"
#include "vdev_iterator.h"
#include "zfsd_event.h"
#include "case_file.h"
#include "vdev.h"
#include "zfsd.h"
#include "zfsd_exception.h"
#include "zpool_list.h"
__FBSDID("$FreeBSD$");
/*============================ Namespace Control =============================*/
using DevdCtl::Event;
using DevdCtl::Guid;
using DevdCtl::NVPairMap;
using std::stringstream;
/*=========================== Class Implementations ==========================*/
/*-------------------------------- DevfsEvent --------------------------------*/
//- DevfsEvent Static Public Methods -------------------------------------------
Event *
DevfsEvent::Builder(Event::Type type,
NVPairMap &nvPairs,
const string &eventString)
{
return (new DevfsEvent(type, nvPairs, eventString));
}
//- DevfsEvent Static Protected Methods ----------------------------------------
nvlist_t *
DevfsEvent::ReadLabel(int devFd, bool &inUse, bool &degraded)
{
pool_state_t poolState;
char *poolName;
boolean_t b_inuse;
inUse = false;
degraded = false;
poolName = NULL;
if (zpool_in_use(g_zfsHandle, devFd, &poolState,
&poolName, &b_inuse) == 0) {
nvlist_t *devLabel;
inUse = b_inuse == B_TRUE;
if (poolName != NULL)
free(poolName);
if (zpool_read_label(devFd, &devLabel) != 0
|| devLabel == NULL)
return (NULL);
try {
Vdev vdev(devLabel);
degraded = vdev.State() != VDEV_STATE_HEALTHY;
return (devLabel);
} catch (ZfsdException &exp) {
string devName = fdevname(devFd);
string devPath = _PATH_DEV + devName;
string context("DevfsEvent::ReadLabel: "
+ devPath + ": ");
exp.GetString().insert(0, context);
exp.Log();
}
}
return (NULL);
}
bool
DevfsEvent::OnlineByLabel(const string &devPath, const string& physPath,
nvlist_t *devConfig)
{
try {
/*
* A device with ZFS label information has been
* inserted. If it matches a device for which we
* have a case, see if we can solve that case.
*/
syslog(LOG_INFO, "Interrogating VDEV label for %s\n",
devPath.c_str());
Vdev vdev(devConfig);
CaseFile *caseFile(CaseFile::Find(vdev.PoolGUID(),
vdev.GUID()));
if (caseFile != NULL)
return (caseFile->ReEvaluate(devPath, physPath, &vdev));
} catch (ZfsdException &exp) {
string context("DevfsEvent::OnlineByLabel: " + devPath + ": ");
exp.GetString().insert(0, context);
exp.Log();
}
return (false);
}
//- DevfsEvent Virtual Public Methods ------------------------------------------
Event *
DevfsEvent::DeepCopy() const
{
return (new DevfsEvent(*this));
}
bool
DevfsEvent::Process() const
{
/*
* We are only concerned with newly discovered
* devices that can be ZFS vdevs.
*/
if (Value("type") != "CREATE" || !IsDiskDev())
return (false);
/* Log the event since it is of interest. */
Log(LOG_INFO);
string devPath;
if (!DevPath(devPath))
return (false);
int devFd(open(devPath.c_str(), O_RDONLY));
if (devFd == -1)
return (false);
bool inUse;
bool degraded;
nvlist_t *devLabel(ReadLabel(devFd, inUse, degraded));
string physPath;
bool havePhysPath(PhysicalPath(physPath));
string devName;
DevName(devName);
close(devFd);
if (inUse && devLabel != NULL) {
OnlineByLabel(devPath, physPath, devLabel);
} else if (degraded) {
syslog(LOG_INFO, "%s is marked degraded. Ignoring "
"as a replace by physical path candidate.\n",
devName.c_str());
} else if (havePhysPath && IsWholeDev()) {
/*
* TODO: attempt to resolve events using every casefile
* that matches this physpath
*/
CaseFile *caseFile(CaseFile::Find(physPath));
if (caseFile != NULL) {
syslog(LOG_INFO,
"Found CaseFile(%s:%s:%s) - ReEvaluating\n",
caseFile->PoolGUIDString().c_str(),
caseFile->VdevGUIDString().c_str(),
zpool_state_to_name(caseFile->VdevState(),
VDEV_AUX_NONE));
caseFile->ReEvaluate(devPath, physPath, /*vdev*/NULL);
}
}
if (devLabel != NULL)
nvlist_free(devLabel);
return (false);
}
//- DevfsEvent Protected Methods -----------------------------------------------
DevfsEvent::DevfsEvent(Event::Type type, NVPairMap &nvpairs,
const string &eventString)
: DevdCtl::DevfsEvent(type, nvpairs, eventString)
{
}
DevfsEvent::DevfsEvent(const DevfsEvent &src)
: DevdCtl::DevfsEvent::DevfsEvent(src)
{
}
/*-------------------------------- GeomEvent --------------------------------*/
//- GeomEvent Static Public Methods -------------------------------------------
Event *
GeomEvent::Builder(Event::Type type,
NVPairMap &nvPairs,
const string &eventString)
{
return (new GeomEvent(type, nvPairs, eventString));
}
//- GeomEvent Virtual Public Methods ------------------------------------------
Event *
GeomEvent::DeepCopy() const
{
return (new GeomEvent(*this));
}
bool
GeomEvent::Process() const
{
/*
* We are only concerned with physical path changes, because those can
* be used to satisfy autoreplace operations
*/
if (Value("type") != "GEOM::physpath" || !IsDiskDev())
return (false);
/* Log the event since it is of interest. */
Log(LOG_INFO);
string devPath;
if (!DevPath(devPath))
return (false);
string physPath;
bool havePhysPath(PhysicalPath(physPath));
string devName;
DevName(devName);
if (havePhysPath) {
/*
* TODO: attempt to resolve events using every casefile
* that matches this physpath
*/
CaseFile *caseFile(CaseFile::Find(physPath));
if (caseFile != NULL) {
syslog(LOG_INFO,
"Found CaseFile(%s:%s:%s) - ReEvaluating\n",
caseFile->PoolGUIDString().c_str(),
caseFile->VdevGUIDString().c_str(),
zpool_state_to_name(caseFile->VdevState(),
VDEV_AUX_NONE));
caseFile->ReEvaluate(devPath, physPath, /*vdev*/NULL);
}
}
return (false);
}
//- GeomEvent Protected Methods -----------------------------------------------
GeomEvent::GeomEvent(Event::Type type, NVPairMap &nvpairs,
const string &eventString)
: DevdCtl::GeomEvent(type, nvpairs, eventString)
{
}
GeomEvent::GeomEvent(const GeomEvent &src)
: DevdCtl::GeomEvent::GeomEvent(src)
{
}
/*--------------------------------- ZfsEvent ---------------------------------*/
//- ZfsEvent Static Public Methods ---------------------------------------------
DevdCtl::Event *
ZfsEvent::Builder(Event::Type type, NVPairMap &nvpairs,
const string &eventString)
{
return (new ZfsEvent(type, nvpairs, eventString));
}
//- ZfsEvent Virtual Public Methods --------------------------------------------
Event *
ZfsEvent::DeepCopy() const
{
return (new ZfsEvent(*this));
}
bool
ZfsEvent::Process() const
{
string logstr("");
if (!Contains("class") && !Contains("type")) {
syslog(LOG_ERR,
"ZfsEvent::Process: Missing class or type data.");
return (false);
}
/* On config syncs, replay any queued events first. */
if (Value("type").find("misc.fs.zfs.config_sync") == 0) {
/*
* Even if saved events are unconsumed the second time
* around, drop them. Any events that still can't be
* consumed are probably referring to vdevs or pools that
* no longer exist.
*/
ZfsDaemon::Get().ReplayUnconsumedEvents(/*discard*/true);
CaseFile::ReEvaluateByGuid(PoolGUID(), *this);
}
if (Value("type").find("misc.fs.zfs.") == 0) {
/* Configuration changes, resilver events, etc. */
ProcessPoolEvent();
return (false);
}
if (!Contains("pool_guid") || !Contains("vdev_guid")) {
/* Only currently interested in Vdev related events. */
return (false);
}
CaseFile *caseFile(CaseFile::Find(PoolGUID(), VdevGUID()));
if (caseFile != NULL) {
Log(LOG_INFO);
syslog(LOG_INFO, "Evaluating existing case file\n");
caseFile->ReEvaluate(*this);
return (false);
}
/* Skip events that can't be handled. */
Guid poolGUID(PoolGUID());
/* If there are no replicas for a pool, then it's not manageable. */
if (Value("class").find("fs.zfs.vdev.no_replicas") == 0) {
stringstream msg;
msg << "No replicas available for pool " << poolGUID;
msg << ", ignoring";
Log(LOG_INFO);
syslog(LOG_INFO, "%s", msg.str().c_str());
return (false);
}
/*
* Create a case file for this vdev, and have it
* evaluate the event.
*/
ZpoolList zpl(ZpoolList::ZpoolByGUID, &poolGUID);
if (zpl.empty()) {
stringstream msg;
int priority = LOG_INFO;
msg << "ZfsEvent::Process: Event for unknown pool ";
msg << poolGUID << " ";
msg << "queued";
Log(LOG_INFO);
syslog(priority, "%s", msg.str().c_str());
return (true);
}
nvlist_t *vdevConfig = VdevIterator(zpl.front()).Find(VdevGUID());
if (vdevConfig == NULL) {
stringstream msg;
int priority = LOG_INFO;
msg << "ZfsEvent::Process: Event for unknown vdev ";
msg << VdevGUID() << " ";
msg << "queued";
Log(LOG_INFO);
syslog(priority, "%s", msg.str().c_str());
return (true);
}
Vdev vdev(zpl.front(), vdevConfig);
caseFile = &CaseFile::Create(vdev);
if (caseFile->ReEvaluate(*this) == false) {
stringstream msg;
int priority = LOG_INFO;
msg << "ZfsEvent::Process: Unconsumed event for vdev(";
msg << zpool_get_name(zpl.front()) << ",";
msg << vdev.GUID() << ") ";
msg << "queued";
Log(LOG_INFO);
syslog(priority, "%s", msg.str().c_str());
return (true);
}
return (false);
}
//- ZfsEvent Protected Methods -------------------------------------------------
ZfsEvent::ZfsEvent(Event::Type type, NVPairMap &nvpairs,
const string &eventString)
: DevdCtl::ZfsEvent(type, nvpairs, eventString)
{
}
ZfsEvent::ZfsEvent(const ZfsEvent &src)
: DevdCtl::ZfsEvent(src)
{
}
/*
* Sometimes the kernel won't detach a spare when it is no longer needed. This
* can happen for example if a drive is removed, then either the pool is
* exported or the machine is powered off, then the drive is reinserted, then
* the machine is powered on or the pool is imported. ZFSD must detach these
* spares itself.
*/
void
ZfsEvent::CleanupSpares() const
{
Guid poolGUID(PoolGUID());
ZpoolList zpl(ZpoolList::ZpoolByGUID, &poolGUID);
if (!zpl.empty()) {
zpool_handle_t* hdl;
hdl = zpl.front();
VdevIterator(hdl).Each(TryDetach, (void*)hdl);
}
}
void
ZfsEvent::ProcessPoolEvent() const
{
bool degradedDevice(false);
/* The pool is destroyed. Discard any open cases */
if (Value("type") == "misc.fs.zfs.pool_destroy") {
Log(LOG_INFO);
CaseFile::ReEvaluateByGuid(PoolGUID(), *this);
return;
}
CaseFile *caseFile(CaseFile::Find(PoolGUID(), VdevGUID()));
if (caseFile != NULL) {
if (caseFile->VdevState() != VDEV_STATE_UNKNOWN
&& caseFile->VdevState() < VDEV_STATE_HEALTHY)
degradedDevice = true;
Log(LOG_INFO);
caseFile->ReEvaluate(*this);
}
else if (Value("type") == "misc.fs.zfs.resilver_finish")
{
/*
* It's possible to get a resilver_finish event with no
* corresponding casefile. For example, if a damaged pool were
* exported, repaired, then reimported.
*/
Log(LOG_INFO);
CleanupSpares();
}
if (Value("type") == "misc.fs.zfs.vdev_remove"
&& degradedDevice == false) {
/* See if any other cases can make use of this device. */
Log(LOG_INFO);
ZfsDaemon::RequestSystemRescan();
}
}
bool
ZfsEvent::TryDetach(Vdev &vdev, void *cbArg)
{
/*
* Outline:
* if this device is a spare, and its parent includes one healthy,
* non-spare child, then detach this device.
*/
zpool_handle_t *hdl(static_cast<zpool_handle_t*>(cbArg));
if (vdev.IsSpare()) {
std::list<Vdev> siblings;
std::list<Vdev>::iterator siblings_it;
boolean_t cleanup = B_FALSE;
Vdev parent = vdev.Parent();
siblings = parent.Children();
/* Determine whether the parent should be cleaned up */
for (siblings_it = siblings.begin();
siblings_it != siblings.end();
siblings_it++) {
Vdev sibling = *siblings_it;
if (!sibling.IsSpare() &&
sibling.State() == VDEV_STATE_HEALTHY) {
cleanup = B_TRUE;
break;
}
}
if (cleanup) {
syslog(LOG_INFO, "Detaching spare vdev %s from pool %s",
vdev.Path().c_str(), zpool_get_name(hdl));
zpool_vdev_detach(hdl, vdev.Path().c_str());
}
}
/* Always return false, because there may be other spares to detach */
return (false);
}