freebsd-src/sbin/hastd/control.c
Mikolaj Golub 3a0ce7f655 MFC r257155, r257582, r259191, r259192, r259193, r259194, r259195, r259196:
r257155:

Make hastctl list command output current queue sizes.

Reviewed by:	pjd

r257582 (pjd):

Correct alignment.

r259191:

For memsync replication, hio_countdown is used not only as an
indication when a request can be moved to done queue, but also for
detecting the current state of memsync request.

This approach has problems, e.g. leaking a request if memsynk ack from
the secondary failed, or racy usage of write_complete, which should be
called only once per write request, but for memsync can be entered by
local_send_thread and ggate_send_thread simultaneously.

So the following approach is implemented instead:

1) Use hio_countdown only for counting components we waiting to
   complete, i.e. initially it is always 2 for any replication mode.

2) To distinguish between "memsync ack" and "memsync fin" responses
   from the secondary, add and use hio_memsyncacked field.

3) write_complete() in component threads is called only before
   releasing hio_countdown (i.e. before the hio may be returned to the
   done queue).

4) Add and use hio_writecount refcounter to detect when
   write_complete() can be called in memsync case.

Reported by:	Pete French petefrench ingresso.co.uk
Tested by:	Pete French petefrench ingresso.co.uk

r259192:

Add some macros to make the code more readable (no functional chages).

r259193:

Fix compiler warnings.

r259194:

In remote_send_thread, if sending a request fails don't take the
request back from the receive queue -- it might already be processed
by remote_recv_thread, which lead to crashes like below:

  (primary) Unable to receive reply header: Connection reset by peer.
  (primary) Unable to send request (Connection reset by peer):
      WRITE(954662912, 131072).
  (primary) Disconnected from kopusha:7772.
  (primary) Increasing localcnt to 1.
  (primary) Assertion failed: (old > 0), function refcnt_release,
      file refcnt.h, line 62.

Taking the request back was not necessary (it would properly be
processed by the remote_recv_thread) and only complicated things.

r259195:

Send wakeup to threads waiting on empty queue before releasing the
lock to decrease spurious wakeups.

Submitted by:	davidxu

r259196:

Check remote protocol version only for the first connection (when it
is actually sent by the remote node).

Otherwise it generated confusing "Negotiated protocol version 1" debug
messages when processing the second connection.
2013-12-28 19:21:22 +00:00

523 lines
14 KiB
C

/*-
* Copyright (c) 2009-2010 The FreeBSD Foundation
* All rights reserved.
*
* This software was developed by Pawel Jakub Dawidek under sponsorship from
* the FreeBSD Foundation.
*
* 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 AUTHORS 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 AUTHORS 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/types.h>
#include <sys/wait.h>
#include <errno.h>
#include <pthread.h>
#include <signal.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include "hast.h"
#include "hastd.h"
#include "hast_checksum.h"
#include "hast_compression.h"
#include "hast_proto.h"
#include "hooks.h"
#include "nv.h"
#include "pjdlog.h"
#include "proto.h"
#include "subr.h"
#include "control.h"
void
child_cleanup(struct hast_resource *res)
{
proto_close(res->hr_ctrl);
res->hr_ctrl = NULL;
if (res->hr_event != NULL) {
proto_close(res->hr_event);
res->hr_event = NULL;
}
if (res->hr_conn != NULL) {
proto_close(res->hr_conn);
res->hr_conn = NULL;
}
res->hr_workerpid = 0;
}
static void
control_set_role_common(struct hastd_config *cfg, struct nv *nvout,
uint8_t role, struct hast_resource *res, const char *name, unsigned int no)
{
int oldrole;
/* Name is always needed. */
if (name != NULL)
nv_add_string(nvout, name, "resource%u", no);
if (res == NULL) {
PJDLOG_ASSERT(cfg != NULL);
PJDLOG_ASSERT(name != NULL);
TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) {
if (strcmp(res->hr_name, name) == 0)
break;
}
if (res == NULL) {
nv_add_int16(nvout, EHAST_NOENTRY, "error%u", no);
return;
}
}
PJDLOG_ASSERT(res != NULL);
/* Send previous role back. */
nv_add_string(nvout, role2str(res->hr_role), "role%u", no);
/* Nothing changed, return here. */
if (role == res->hr_role)
return;
pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role));
pjdlog_info("Role changed to %s.", role2str(role));
/* Change role to the new one. */
oldrole = res->hr_role;
res->hr_role = role;
pjdlog_prefix_set("[%s] (%s) ", res->hr_name, role2str(res->hr_role));
/*
* If previous role was primary or secondary we have to kill process
* doing that work.
*/
if (res->hr_workerpid != 0) {
if (kill(res->hr_workerpid, SIGTERM) == -1) {
pjdlog_errno(LOG_WARNING,
"Unable to kill worker process %u",
(unsigned int)res->hr_workerpid);
} else if (waitpid(res->hr_workerpid, NULL, 0) !=
res->hr_workerpid) {
pjdlog_errno(LOG_WARNING,
"Error while waiting for worker process %u",
(unsigned int)res->hr_workerpid);
} else {
pjdlog_debug(1, "Worker process %u stopped.",
(unsigned int)res->hr_workerpid);
}
child_cleanup(res);
}
/* Start worker process if we are changing to primary. */
if (role == HAST_ROLE_PRIMARY)
hastd_primary(res);
pjdlog_prefix_set("%s", "");
hook_exec(res->hr_exec, "role", res->hr_name, role2str(oldrole),
role2str(res->hr_role), NULL);
}
void
control_set_role(struct hast_resource *res, uint8_t role)
{
control_set_role_common(NULL, NULL, role, res, NULL, 0);
}
static void
control_status_worker(struct hast_resource *res, struct nv *nvout,
unsigned int no)
{
struct nv *cnvin, *cnvout;
const char *str;
int error;
cnvin = NULL;
/*
* Prepare and send command to worker process.
*/
cnvout = nv_alloc();
nv_add_uint8(cnvout, CONTROL_STATUS, "cmd");
error = nv_error(cnvout);
if (error != 0) {
pjdlog_common(LOG_ERR, 0, error,
"Unable to prepare control header");
goto end;
}
if (hast_proto_send(res, res->hr_ctrl, cnvout, NULL, 0) == -1) {
error = errno;
pjdlog_errno(LOG_ERR, "Unable to send control header");
goto end;
}
/*
* Receive response.
*/
if (hast_proto_recv_hdr(res->hr_ctrl, &cnvin) == -1) {
error = errno;
pjdlog_errno(LOG_ERR, "Unable to receive control header");
goto end;
}
error = nv_get_int16(cnvin, "error");
if (error != 0)
goto end;
if ((str = nv_get_string(cnvin, "status")) == NULL) {
error = ENOENT;
pjdlog_errno(LOG_ERR, "Field 'status' is missing.");
goto end;
}
nv_add_string(nvout, str, "status%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "dirty"), "dirty%u", no);
nv_add_uint32(nvout, nv_get_uint32(cnvin, "extentsize"),
"extentsize%u", no);
nv_add_uint32(nvout, nv_get_uint32(cnvin, "keepdirty"),
"keepdirty%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_read"),
"stat_read%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_write"),
"stat_write%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_delete"),
"stat_delete%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_flush"),
"stat_flush%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_activemap_update"),
"stat_activemap_update%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_read_error"),
"stat_read_error%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_write_error"),
"stat_write_error%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_delete_error"),
"stat_delete_error%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "stat_flush_error"),
"stat_flush_error%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "idle_queue_size"),
"idle_queue_size%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "local_queue_size"),
"local_queue_size%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "send_queue_size"),
"send_queue_size%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "recv_queue_size"),
"recv_queue_size%u", no);
nv_add_uint64(nvout, nv_get_uint64(cnvin, "done_queue_size"),
"done_queue_size%u", no);
end:
if (cnvin != NULL)
nv_free(cnvin);
if (cnvout != NULL)
nv_free(cnvout);
if (error != 0)
nv_add_int16(nvout, error, "error");
}
static void
control_status(struct hastd_config *cfg, struct nv *nvout,
struct hast_resource *res, const char *name, unsigned int no)
{
PJDLOG_ASSERT(cfg != NULL);
PJDLOG_ASSERT(nvout != NULL);
PJDLOG_ASSERT(name != NULL);
/* Name is always needed. */
nv_add_string(nvout, name, "resource%u", no);
if (res == NULL) {
TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) {
if (strcmp(res->hr_name, name) == 0)
break;
}
if (res == NULL) {
nv_add_int16(nvout, EHAST_NOENTRY, "error%u", no);
return;
}
}
PJDLOG_ASSERT(res != NULL);
nv_add_string(nvout, res->hr_provname, "provname%u", no);
nv_add_string(nvout, res->hr_localpath, "localpath%u", no);
nv_add_string(nvout, res->hr_remoteaddr, "remoteaddr%u", no);
if (res->hr_sourceaddr[0] != '\0')
nv_add_string(nvout, res->hr_sourceaddr, "sourceaddr%u", no);
switch (res->hr_replication) {
case HAST_REPLICATION_FULLSYNC:
nv_add_string(nvout, "fullsync", "replication%u", no);
break;
case HAST_REPLICATION_MEMSYNC:
nv_add_string(nvout, "memsync", "replication%u", no);
break;
case HAST_REPLICATION_ASYNC:
nv_add_string(nvout, "async", "replication%u", no);
break;
default:
nv_add_string(nvout, "unknown", "replication%u", no);
break;
}
nv_add_string(nvout, checksum_name(res->hr_checksum),
"checksum%u", no);
nv_add_string(nvout, compression_name(res->hr_compression),
"compression%u", no);
nv_add_string(nvout, role2str(res->hr_role), "role%u", no);
nv_add_int32(nvout, res->hr_workerpid, "workerpid%u", no);
switch (res->hr_role) {
case HAST_ROLE_PRIMARY:
PJDLOG_ASSERT(res->hr_workerpid != 0);
/* FALLTHROUGH */
case HAST_ROLE_SECONDARY:
if (res->hr_workerpid != 0)
break;
/* FALLTHROUGH */
default:
return;
}
/*
* If we are here, it means that we have a worker process, which we
* want to ask some questions.
*/
control_status_worker(res, nvout, no);
}
void
control_handle(struct hastd_config *cfg)
{
struct proto_conn *conn;
struct nv *nvin, *nvout;
unsigned int ii;
const char *str;
uint8_t cmd, role;
int error;
if (proto_accept(cfg->hc_controlconn, &conn) == -1) {
pjdlog_errno(LOG_ERR, "Unable to accept control connection");
return;
}
cfg->hc_controlin = conn;
nvin = nvout = NULL;
role = HAST_ROLE_UNDEF;
if (hast_proto_recv_hdr(conn, &nvin) == -1) {
pjdlog_errno(LOG_ERR, "Unable to receive control header");
nvin = NULL;
goto close;
}
/* Obtain command code. 0 means that nv_get_uint8() failed. */
cmd = nv_get_uint8(nvin, "cmd");
if (cmd == 0) {
pjdlog_error("Control header is missing 'cmd' field.");
goto close;
}
/* Allocate outgoing nv structure. */
nvout = nv_alloc();
if (nvout == NULL) {
pjdlog_error("Unable to allocate header for control response.");
goto close;
}
error = 0;
str = nv_get_string(nvin, "resource0");
if (str == NULL) {
pjdlog_error("Control header is missing 'resource0' field.");
error = EHAST_INVALID;
goto fail;
}
if (cmd == HASTCTL_CMD_SETROLE) {
role = nv_get_uint8(nvin, "role");
switch (role) {
case HAST_ROLE_INIT:
case HAST_ROLE_PRIMARY:
case HAST_ROLE_SECONDARY:
break;
default:
pjdlog_error("Invalid role received (%hhu).", role);
error = EHAST_INVALID;
goto fail;
}
}
if (strcmp(str, "all") == 0) {
struct hast_resource *res;
/* All configured resources. */
ii = 0;
TAILQ_FOREACH(res, &cfg->hc_resources, hr_next) {
switch (cmd) {
case HASTCTL_CMD_SETROLE:
control_set_role_common(cfg, nvout, role, res,
res->hr_name, ii++);
break;
case HASTCTL_CMD_STATUS:
control_status(cfg, nvout, res, res->hr_name,
ii++);
break;
default:
pjdlog_error("Invalid command received (%hhu).",
cmd);
error = EHAST_UNIMPLEMENTED;
goto fail;
}
}
} else {
/* Only selected resources. */
for (ii = 0; ; ii++) {
str = nv_get_string(nvin, "resource%u", ii);
if (str == NULL)
break;
switch (cmd) {
case HASTCTL_CMD_SETROLE:
control_set_role_common(cfg, nvout, role, NULL,
str, ii);
break;
case HASTCTL_CMD_STATUS:
control_status(cfg, nvout, NULL, str, ii);
break;
default:
pjdlog_error("Invalid command received (%hhu).",
cmd);
error = EHAST_UNIMPLEMENTED;
goto fail;
}
}
}
if (nv_error(nvout) != 0)
goto close;
fail:
if (error != 0)
nv_add_int16(nvout, error, "error");
if (hast_proto_send(NULL, conn, nvout, NULL, 0) == -1)
pjdlog_errno(LOG_ERR, "Unable to send control response");
close:
if (nvin != NULL)
nv_free(nvin);
if (nvout != NULL)
nv_free(nvout);
proto_close(conn);
cfg->hc_controlin = NULL;
}
/*
* Thread handles control requests from the parent.
*/
void *
ctrl_thread(void *arg)
{
struct hast_resource *res = arg;
struct nv *nvin, *nvout;
uint8_t cmd;
for (;;) {
if (hast_proto_recv_hdr(res->hr_ctrl, &nvin) == -1) {
if (sigexit_received)
pthread_exit(NULL);
pjdlog_errno(LOG_ERR,
"Unable to receive control message");
kill(getpid(), SIGTERM);
pthread_exit(NULL);
}
cmd = nv_get_uint8(nvin, "cmd");
if (cmd == 0) {
pjdlog_error("Control message is missing 'cmd' field.");
nv_free(nvin);
continue;
}
nvout = nv_alloc();
switch (cmd) {
case CONTROL_STATUS:
if (res->hr_remotein != NULL &&
res->hr_remoteout != NULL) {
nv_add_string(nvout, "complete", "status");
} else {
nv_add_string(nvout, "degraded", "status");
}
nv_add_uint32(nvout, (uint32_t)res->hr_extentsize,
"extentsize");
if (res->hr_role == HAST_ROLE_PRIMARY) {
nv_add_uint32(nvout,
(uint32_t)res->hr_keepdirty, "keepdirty");
nv_add_uint64(nvout,
(uint64_t)(activemap_ndirty(res->hr_amp) *
res->hr_extentsize), "dirty");
} else {
nv_add_uint32(nvout, (uint32_t)0, "keepdirty");
nv_add_uint64(nvout, (uint64_t)0, "dirty");
}
nv_add_uint64(nvout, res->hr_stat_read, "stat_read");
nv_add_uint64(nvout, res->hr_stat_write, "stat_write");
nv_add_uint64(nvout, res->hr_stat_delete,
"stat_delete");
nv_add_uint64(nvout, res->hr_stat_flush, "stat_flush");
nv_add_uint64(nvout, res->hr_stat_activemap_update,
"stat_activemap_update");
nv_add_uint64(nvout, res->hr_stat_read_error,
"stat_read_error");
nv_add_uint64(nvout, res->hr_stat_write_error +
res->hr_stat_activemap_write_error,
"stat_write_error");
nv_add_uint64(nvout, res->hr_stat_delete_error,
"stat_delete_error");
nv_add_uint64(nvout, res->hr_stat_flush_error +
res->hr_stat_activemap_flush_error,
"stat_flush_error");
res->output_status_aux(nvout);
nv_add_int16(nvout, 0, "error");
break;
case CONTROL_RELOAD:
/*
* When parent receives SIGHUP and discovers that
* something related to us has changes, it sends reload
* message to us.
*/
PJDLOG_ASSERT(res->hr_role == HAST_ROLE_PRIMARY);
primary_config_reload(res, nvin);
nv_add_int16(nvout, 0, "error");
break;
default:
nv_add_int16(nvout, EINVAL, "error");
break;
}
nv_free(nvin);
if (nv_error(nvout) != 0) {
pjdlog_error("Unable to create answer on control message.");
nv_free(nvout);
continue;
}
if (hast_proto_send(NULL, res->hr_ctrl, nvout, NULL, 0) == -1) {
pjdlog_errno(LOG_ERR,
"Unable to send reply to control message");
}
nv_free(nvout);
}
/* NOTREACHED */
return (NULL);
}