nbd: keep send_mutex/free_sema handling outside nbd_co_do_establish_connection

Elevate s->in_flight early so that other incoming requests will wait
on the CoQueue in nbd_co_send_request; restart them after getting back
from nbd_reconnect_attempt.  This could be after the reconnect timer or
nbd_cancel_in_flight have cancelled the attempt, so there is no
need anymore to cancel the requests there.

nbd_co_send_request now handles both stopping and restarting pending
requests after a successful connection, and there is no need to
hold send_mutex in nbd_co_do_establish_connection.  The current setup
is confusing because nbd_co_do_establish_connection is called both with
send_mutex taken and without it.  Before the patch it uses free_sema which
(at least in theory...) is protected by send_mutex, after the patch it
does not anymore.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220414175756.671165-5-pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: wrap long line]
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@openvz.org>
Reviewed-by: Lukas Straub <lukasstraub2@web.de>
Signed-off-by: Eric Blake <eblake@redhat.com>
This commit is contained in:
Paolo Bonzini 2022-04-14 19:57:51 +02:00 committed by Eric Blake
parent 172f5f1a40
commit 8610b4491f
2 changed files with 34 additions and 37 deletions

View File

@ -59,7 +59,8 @@ int coroutine_fn bdrv_co_writev_vmstate(BlockDriverState *bs,
QEMUIOVector *qiov, int64_t pos);
int coroutine_fn
nbd_co_do_establish_connection(BlockDriverState *bs, Error **errp);
nbd_co_do_establish_connection(BlockDriverState *bs, bool blocking,
Error **errp);
int coroutine_fn
@ -109,7 +110,7 @@ bdrv_common_block_status_above(BlockDriverState *bs,
BlockDriverState **file,
int *depth);
int generated_co_wrapper
nbd_do_establish_connection(BlockDriverState *bs, Error **errp);
nbd_do_establish_connection(BlockDriverState *bs, bool blocking, Error **errp);
int generated_co_wrapper
blk_do_preadv(BlockBackend *blk, int64_t offset, int64_t bytes,

View File

@ -187,9 +187,6 @@ static void reconnect_delay_timer_cb(void *opaque)
if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT) {
s->state = NBD_CLIENT_CONNECTING_NOWAIT;
nbd_co_establish_connection_cancel(s->conn);
while (qemu_co_enter_next(&s->free_sema, NULL)) {
/* Resume all queued requests */
}
}
reconnect_delay_timer_del(s);
@ -310,11 +307,10 @@ static int nbd_handle_updated_info(BlockDriverState *bs, Error **errp)
}
int coroutine_fn nbd_co_do_establish_connection(BlockDriverState *bs,
Error **errp)
bool blocking, Error **errp)
{
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
int ret;
bool blocking = nbd_client_connecting_wait(s);
IO_CODE();
assert(!s->ioc);
@ -350,7 +346,6 @@ int coroutine_fn nbd_co_do_establish_connection(BlockDriverState *bs,
/* successfully connected */
s->state = NBD_CLIENT_CONNECTED;
qemu_co_queue_restart_all(&s->free_sema);
return 0;
}
@ -358,25 +353,25 @@ int coroutine_fn nbd_co_do_establish_connection(BlockDriverState *bs,
/* called under s->send_mutex */
static coroutine_fn void nbd_reconnect_attempt(BDRVNBDState *s)
{
assert(nbd_client_connecting(s));
assert(s->in_flight == 0);
if (nbd_client_connecting_wait(s) && s->reconnect_delay &&
!s->reconnect_delay_timer)
{
/*
* It's first reconnect attempt after switching to
* NBD_CLIENT_CONNECTING_WAIT
*/
reconnect_delay_timer_init(s,
qemu_clock_get_ns(QEMU_CLOCK_REALTIME) +
s->reconnect_delay * NANOSECONDS_PER_SECOND);
}
bool blocking = nbd_client_connecting_wait(s);
/*
* Now we are sure that nobody is accessing the channel, and no one will
* try until we set the state to CONNECTED.
*/
assert(nbd_client_connecting(s));
assert(s->in_flight == 1);
if (blocking && !s->reconnect_delay_timer) {
/*
* It's the first reconnect attempt after switching to
* NBD_CLIENT_CONNECTING_WAIT
*/
g_assert(s->reconnect_delay);
reconnect_delay_timer_init(s,
qemu_clock_get_ns(QEMU_CLOCK_REALTIME) +
s->reconnect_delay * NANOSECONDS_PER_SECOND);
}
/* Finalize previous connection if any */
if (s->ioc) {
@ -387,7 +382,9 @@ static coroutine_fn void nbd_reconnect_attempt(BDRVNBDState *s)
s->ioc = NULL;
}
nbd_co_do_establish_connection(s->bs, NULL);
qemu_co_mutex_unlock(&s->send_mutex);
nbd_co_do_establish_connection(s->bs, blocking, NULL);
qemu_co_mutex_lock(&s->send_mutex);
/*
* The reconnect attempt is done (maybe successfully, maybe not), so
@ -472,21 +469,21 @@ static int coroutine_fn nbd_co_send_request(BlockDriverState *bs,
qemu_co_mutex_lock(&s->send_mutex);
while (s->in_flight == MAX_NBD_REQUESTS ||
(!nbd_client_connected(s) && s->in_flight > 0))
{
(!nbd_client_connected(s) && s->in_flight > 0)) {
qemu_co_queue_wait(&s->free_sema, &s->send_mutex);
}
if (nbd_client_connecting(s)) {
nbd_reconnect_attempt(s);
}
if (!nbd_client_connected(s)) {
rc = -EIO;
goto err;
}
s->in_flight++;
if (!nbd_client_connected(s)) {
if (nbd_client_connecting(s)) {
nbd_reconnect_attempt(s);
qemu_co_queue_restart_all(&s->free_sema);
}
if (!nbd_client_connected(s)) {
rc = -EIO;
goto err;
}
}
for (i = 0; i < MAX_NBD_REQUESTS; i++) {
if (s->requests[i].coroutine == NULL) {
@ -526,8 +523,8 @@ err:
nbd_channel_error(s, rc);
if (i != -1) {
s->requests[i].coroutine = NULL;
s->in_flight--;
}
s->in_flight--;
qemu_co_queue_next(&s->free_sema);
}
qemu_co_mutex_unlock(&s->send_mutex);
@ -1882,7 +1879,7 @@ static int nbd_open(BlockDriverState *bs, QDict *options, int flags,
}
s->state = NBD_CLIENT_CONNECTING_WAIT;
ret = nbd_do_establish_connection(bs, errp);
ret = nbd_do_establish_connection(bs, true, errp);
if (ret < 0) {
goto fail;
}
@ -2048,7 +2045,6 @@ static void nbd_cancel_in_flight(BlockDriverState *bs)
if (s->state == NBD_CLIENT_CONNECTING_WAIT) {
s->state = NBD_CLIENT_CONNECTING_NOWAIT;
qemu_co_queue_restart_all(&s->free_sema);
}
nbd_co_establish_connection_cancel(s->conn);