2008-07-03 15:41:03 +02:00
|
|
|
/*
|
|
|
|
* QEMU Block driver for NBD
|
|
|
|
*
|
2019-10-09 10:41:57 +02:00
|
|
|
* Copyright (c) 2019 Virtuozzo International GmbH.
|
2019-06-11 12:27:19 +02:00
|
|
|
* Copyright (C) 2016 Red Hat, Inc.
|
2008-07-03 15:41:03 +02:00
|
|
|
* Copyright (C) 2008 Bull S.A.S.
|
2008-07-08 20:57:05 +02:00
|
|
|
* Author: Laurent Vivier <Laurent.Vivier@bull.net>
|
2008-07-03 15:41:03 +02:00
|
|
|
*
|
|
|
|
* Some parts:
|
|
|
|
* Copyright (C) 2007 Anthony Liguori <anthony@codemonkey.ws>
|
|
|
|
*
|
|
|
|
* Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
|
|
* of this software and associated documentation files (the "Software"), to deal
|
|
|
|
* in the Software without restriction, including without limitation the rights
|
|
|
|
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
|
|
* copies of the Software, and to permit persons to whom the Software is
|
|
|
|
* furnished to do so, subject to the following conditions:
|
|
|
|
*
|
|
|
|
* The above copyright notice and this permission notice shall be included in
|
|
|
|
* all copies or substantial portions of the Software.
|
|
|
|
*
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
|
|
|
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
|
|
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
|
|
|
* THE SOFTWARE.
|
|
|
|
*/
|
|
|
|
|
2016-01-18 19:01:42 +01:00
|
|
|
#include "qemu/osdep.h"
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
#include "trace.h"
|
2012-12-17 18:20:00 +01:00
|
|
|
#include "qemu/uri.h"
|
2018-02-01 12:18:46 +01:00
|
|
|
#include "qemu/option.h"
|
2019-06-11 12:27:19 +02:00
|
|
|
#include "qemu/cutils.h"
|
Include qemu/main-loop.h less
In my "build everything" tree, changing qemu/main-loop.h triggers a
recompile of some 5600 out of 6600 objects (not counting tests and
objects that don't depend on qemu/osdep.h). It includes block/aio.h,
which in turn includes qemu/event_notifier.h, qemu/notify.h,
qemu/processor.h, qemu/qsp.h, qemu/queue.h, qemu/thread-posix.h,
qemu/thread.h, qemu/timer.h, and a few more.
Include qemu/main-loop.h only where it's needed. Touching it now
recompiles only some 1700 objects. For block/aio.h and
qemu/event_notifier.h, these numbers drop from 5600 to 2800. For the
others, they shrink only slightly.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190812052359.30071-21-armbru@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
2019-08-12 07:23:50 +02:00
|
|
|
#include "qemu/main-loop.h"
|
2020-12-28 16:08:44 +01:00
|
|
|
#include "qemu/atomic.h"
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2018-02-11 10:36:01 +01:00
|
|
|
#include "qapi/qapi-visit-sockets.h"
|
2014-07-18 20:24:59 +02:00
|
|
|
#include "qapi/qmp/qstring.h"
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
#include "qapi/clone-visitor.h"
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
#include "block/qdict.h"
|
|
|
|
#include "block/nbd.h"
|
|
|
|
#include "block/block_int.h"
|
2008-07-03 15:41:03 +02:00
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
#include "qemu/yank.h"
|
|
|
|
|
2010-08-25 22:48:33 +02:00
|
|
|
#define EN_OPTSTR ":exportname="
|
2019-06-11 12:27:19 +02:00
|
|
|
#define MAX_NBD_REQUESTS 16
|
|
|
|
|
|
|
|
#define HANDLE_TO_INDEX(bs, handle) ((handle) ^ (uint64_t)(intptr_t)(bs))
|
|
|
|
#define INDEX_TO_HANDLE(bs, index) ((index) ^ (uint64_t)(intptr_t)(bs))
|
|
|
|
|
|
|
|
typedef struct {
|
|
|
|
Coroutine *coroutine;
|
|
|
|
uint64_t offset; /* original offset of the request */
|
|
|
|
bool receiving; /* waiting for connection_co? */
|
|
|
|
} NBDClientRequest;
|
|
|
|
|
2019-06-18 13:43:22 +02:00
|
|
|
typedef enum NBDClientState {
|
2019-10-09 10:41:57 +02:00
|
|
|
NBD_CLIENT_CONNECTING_WAIT,
|
|
|
|
NBD_CLIENT_CONNECTING_NOWAIT,
|
2019-06-18 13:43:22 +02:00
|
|
|
NBD_CLIENT_CONNECTED,
|
|
|
|
NBD_CLIENT_QUIT
|
|
|
|
} NBDClientState;
|
|
|
|
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
typedef struct NBDConnectThread {
|
|
|
|
/* Initialization constants */
|
|
|
|
SocketAddress *saddr; /* address to connect to */
|
|
|
|
|
2021-06-10 12:07:41 +02:00
|
|
|
QemuMutex mutex;
|
|
|
|
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
/*
|
2021-06-10 12:07:41 +02:00
|
|
|
* @sioc and @err represent a connection attempt. While running
|
|
|
|
* is true, they are only used by the connection thread, and mutex
|
|
|
|
* locking is not needed. Once the thread finishes,
|
|
|
|
* nbd_co_establish_connection then steals these pointers while
|
|
|
|
* under the mutex.
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
*/
|
|
|
|
QIOChannelSocket *sioc;
|
|
|
|
Error *err;
|
|
|
|
|
2021-06-10 12:07:41 +02:00
|
|
|
/* All further fields are accessed only under mutex */
|
|
|
|
bool running; /* thread is running now */
|
|
|
|
bool detached; /* thread is detached and should cleanup the state */
|
block/nbd: simplify waking of nbd_co_establish_connection()
Instead of managing connect_bh, bh_ctx, and wait_connect fields, we
can use a single link to the waiting coroutine with proper mutex
protection.
So new logic is:
nbd_co_establish_connection() sets wait_co under the mutex, releases
the mutex, then yield()s. Note that wait_co may be scheduled by the
thread immediately after unlocking the mutex. Still, the main thread
(or iothread) will not reach the code for entering the coroutine until
the yield(), so we are safe.
connect_thread_func() and nbd_co_establish_connection_cancel() do
the following to handle wait_co:
Under the mutex, if thr->wait_co is not NULL, make it NULL and
schedule it. This way, we avoid scheduling the coroutine twice.
Still scheduling is a bit different:
In connect_thread_func() we can just call aio_co_wake under mutex,
after commit
[async: the main AioContext is only "current" if under the BQL]
we are sure that aio_co_wake() will not try to acquire the aio context
and do qemu_aio_coroutine_enter() but simply schedule the coroutine by
aio_co_schedule().
nbd_co_establish_connection_cancel() will be called from non-coroutine
context in further patch and will be able to go through
qemu_aio_coroutine_enter() path of aio_co_wake(). So keep current
behavior of waking the coroutine after the critical section.
Also, this commit reduces the dependence of
nbd_co_establish_connection() on the internals of bs (we now use a
generic pointer to the coroutine, instead of direct use of
s->connection_co). This is a step towards splitting the connection
API out of nbd.c.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210610100802.5888-11-vsementsov@virtuozzo.com>
Reviewied-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2021-06-10 12:07:40 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* wait_co: if non-NULL, which coroutine to wake in
|
|
|
|
* nbd_co_establish_connection() after yield()
|
|
|
|
*/
|
|
|
|
Coroutine *wait_co;
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
} NBDConnectThread;
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
typedef struct BDRVNBDState {
|
2019-06-11 12:27:19 +02:00
|
|
|
QIOChannelSocket *sioc; /* The master data channel */
|
|
|
|
QIOChannel *ioc; /* The current I/O channel which may differ (eg TLS) */
|
|
|
|
NBDExportInfo info;
|
|
|
|
|
|
|
|
CoMutex send_mutex;
|
|
|
|
CoQueue free_sema;
|
|
|
|
Coroutine *connection_co;
|
2020-01-22 17:45:28 +01:00
|
|
|
Coroutine *teardown_co;
|
2021-05-17 12:05:47 +02:00
|
|
|
QemuCoSleep reconnect_sleep;
|
2019-10-09 10:41:57 +02:00
|
|
|
bool drained;
|
|
|
|
bool wait_drained_end;
|
2019-06-11 12:27:19 +02:00
|
|
|
int in_flight;
|
2019-06-18 13:43:22 +02:00
|
|
|
NBDClientState state;
|
2019-10-09 10:41:57 +02:00
|
|
|
bool wait_in_flight;
|
2019-06-11 12:27:19 +02:00
|
|
|
|
block/nbd: fix reconnect-delay
reconnect-delay has a design flaw: we handle it in the same loop where
we do connection attempt. So, reconnect-delay may be exceeded by
unpredictable time of connection attempt.
Let's instead use separate timer.
How to reproduce the bug:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. On node2 start qemu-io:
./build/qemu-io --image-opts \
driver=nbd,server.type=inet,server.host=192.168.100.5,server.port=10809,reconnect-delay=15
4. Type 'read 0 512' in qemu-io interface to check that connection
works
Be careful: you should make steps 5-7 in a short time, less than 15
seconds.
5. Kill nbd server on node1
6. Run 'read 0 512' in qemu-io interface again, to be sure that nbd
client goes to reconnect loop.
7. On node1 run the following command
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
This will make the connect() call of qemu-io at node2 take a long time.
And you'll see that read command in qemu-io will hang for a long time,
more than 15 seconds specified by reconnect-delay parameter. It's the
bug.
8. Don't forget to drop iptables rule on node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Important note: Step [5] is necessary to reproduce _this_ bug. If we
miss step [5], the read command (step 6) will hang for a long time and
this commit doesn't help, because there will be not long connect() to
unreachable host, but long sendmsg() to unreachable host, which should
be fixed by enabling and adjusting keep-alive on the socket, which is a
thing for further patch set.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200903190301.367620-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-09-03 21:03:00 +02:00
|
|
|
QEMUTimer *reconnect_delay_timer;
|
|
|
|
|
2019-06-11 12:27:19 +02:00
|
|
|
NBDClientRequest requests[MAX_NBD_REQUESTS];
|
|
|
|
NBDReply reply;
|
|
|
|
BlockDriverState *bs;
|
2016-08-15 15:29:26 +02:00
|
|
|
|
2019-06-18 13:43:24 +02:00
|
|
|
/* Connection parameters */
|
|
|
|
uint32_t reconnect_delay;
|
2017-04-26 09:36:40 +02:00
|
|
|
SocketAddress *saddr;
|
2016-10-25 15:11:34 +02:00
|
|
|
char *export, *tlscredsid;
|
2019-06-18 13:43:24 +02:00
|
|
|
QCryptoTLSCreds *tlscreds;
|
|
|
|
const char *hostname;
|
|
|
|
char *x_dirty_bitmap;
|
2020-10-27 06:05:55 +01:00
|
|
|
bool alloc_depth;
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
|
|
|
|
NBDConnectThread *connect_thread;
|
2008-07-03 15:41:03 +02:00
|
|
|
} BDRVNBDState;
|
|
|
|
|
2021-06-10 12:07:37 +02:00
|
|
|
static void nbd_free_connect_thread(NBDConnectThread *thr);
|
2020-12-28 16:08:44 +01:00
|
|
|
static int nbd_establish_connection(BlockDriverState *bs, SocketAddress *saddr,
|
|
|
|
Error **errp);
|
2021-06-10 12:07:42 +02:00
|
|
|
static coroutine_fn QIOChannelSocket *
|
|
|
|
nbd_co_establish_connection(NBDConnectThread *thr, Error **errp);
|
2021-06-10 12:07:37 +02:00
|
|
|
static void nbd_co_establish_connection_cancel(BlockDriverState *bs);
|
2020-12-28 16:08:44 +01:00
|
|
|
static int nbd_client_handshake(BlockDriverState *bs, Error **errp);
|
|
|
|
static void nbd_yank(void *opaque);
|
2019-10-09 10:41:57 +02:00
|
|
|
|
2021-06-10 12:07:33 +02:00
|
|
|
static void nbd_clear_bdrvstate(BlockDriverState *bs)
|
2019-12-05 04:45:27 +01:00
|
|
|
{
|
2021-06-10 12:07:33 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2021-06-10 12:07:37 +02:00
|
|
|
NBDConnectThread *thr = s->connect_thread;
|
2021-06-10 12:07:41 +02:00
|
|
|
bool do_free = false;
|
2021-06-10 12:07:37 +02:00
|
|
|
|
|
|
|
qemu_mutex_lock(&thr->mutex);
|
2021-06-10 12:07:41 +02:00
|
|
|
assert(!thr->detached);
|
|
|
|
if (thr->running) {
|
|
|
|
thr->detached = true;
|
|
|
|
} else {
|
|
|
|
do_free = true;
|
2021-06-10 12:07:37 +02:00
|
|
|
}
|
|
|
|
qemu_mutex_unlock(&thr->mutex);
|
|
|
|
|
|
|
|
/* the runaway thread will clean up itself */
|
2021-06-10 12:07:41 +02:00
|
|
|
if (do_free) {
|
2021-06-10 12:07:37 +02:00
|
|
|
nbd_free_connect_thread(thr);
|
|
|
|
}
|
2021-06-10 12:07:33 +02:00
|
|
|
|
|
|
|
yank_unregister_instance(BLOCKDEV_YANK_INSTANCE(bs->node_name));
|
|
|
|
|
2019-12-05 04:45:27 +01:00
|
|
|
object_unref(OBJECT(s->tlscreds));
|
|
|
|
qapi_free_SocketAddress(s->saddr);
|
|
|
|
s->saddr = NULL;
|
|
|
|
g_free(s->export);
|
|
|
|
s->export = NULL;
|
|
|
|
g_free(s->tlscredsid);
|
|
|
|
s->tlscredsid = NULL;
|
|
|
|
g_free(s->x_dirty_bitmap);
|
|
|
|
s->x_dirty_bitmap = NULL;
|
|
|
|
}
|
|
|
|
|
2019-06-18 13:43:22 +02:00
|
|
|
static void nbd_channel_error(BDRVNBDState *s, int ret)
|
|
|
|
{
|
2019-10-09 10:41:57 +02:00
|
|
|
if (ret == -EIO) {
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTED) {
|
2019-10-09 10:41:57 +02:00
|
|
|
s->state = s->reconnect_delay ? NBD_CLIENT_CONNECTING_WAIT :
|
|
|
|
NBD_CLIENT_CONNECTING_NOWAIT;
|
|
|
|
}
|
|
|
|
} else {
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTED) {
|
2019-10-09 10:41:57 +02:00
|
|
|
qio_channel_shutdown(s->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL);
|
|
|
|
}
|
|
|
|
s->state = NBD_CLIENT_QUIT;
|
|
|
|
}
|
2019-06-18 13:43:22 +02:00
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
static void nbd_recv_coroutines_wake_all(BDRVNBDState *s)
|
2019-06-11 12:27:19 +02:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < MAX_NBD_REQUESTS; i++) {
|
|
|
|
NBDClientRequest *req = &s->requests[i];
|
|
|
|
|
|
|
|
if (req->coroutine && req->receiving) {
|
|
|
|
aio_co_wake(req->coroutine);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
block/nbd: fix reconnect-delay
reconnect-delay has a design flaw: we handle it in the same loop where
we do connection attempt. So, reconnect-delay may be exceeded by
unpredictable time of connection attempt.
Let's instead use separate timer.
How to reproduce the bug:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. On node2 start qemu-io:
./build/qemu-io --image-opts \
driver=nbd,server.type=inet,server.host=192.168.100.5,server.port=10809,reconnect-delay=15
4. Type 'read 0 512' in qemu-io interface to check that connection
works
Be careful: you should make steps 5-7 in a short time, less than 15
seconds.
5. Kill nbd server on node1
6. Run 'read 0 512' in qemu-io interface again, to be sure that nbd
client goes to reconnect loop.
7. On node1 run the following command
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
This will make the connect() call of qemu-io at node2 take a long time.
And you'll see that read command in qemu-io will hang for a long time,
more than 15 seconds specified by reconnect-delay parameter. It's the
bug.
8. Don't forget to drop iptables rule on node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Important note: Step [5] is necessary to reproduce _this_ bug. If we
miss step [5], the read command (step 6) will hang for a long time and
this commit doesn't help, because there will be not long connect() to
unreachable host, but long sendmsg() to unreachable host, which should
be fixed by enabling and adjusting keep-alive on the socket, which is a
thing for further patch set.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200903190301.367620-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-09-03 21:03:00 +02:00
|
|
|
static void reconnect_delay_timer_del(BDRVNBDState *s)
|
|
|
|
{
|
|
|
|
if (s->reconnect_delay_timer) {
|
|
|
|
timer_free(s->reconnect_delay_timer);
|
|
|
|
s->reconnect_delay_timer = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void reconnect_delay_timer_cb(void *opaque)
|
|
|
|
{
|
|
|
|
BDRVNBDState *s = opaque;
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT) {
|
block/nbd: fix reconnect-delay
reconnect-delay has a design flaw: we handle it in the same loop where
we do connection attempt. So, reconnect-delay may be exceeded by
unpredictable time of connection attempt.
Let's instead use separate timer.
How to reproduce the bug:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. On node2 start qemu-io:
./build/qemu-io --image-opts \
driver=nbd,server.type=inet,server.host=192.168.100.5,server.port=10809,reconnect-delay=15
4. Type 'read 0 512' in qemu-io interface to check that connection
works
Be careful: you should make steps 5-7 in a short time, less than 15
seconds.
5. Kill nbd server on node1
6. Run 'read 0 512' in qemu-io interface again, to be sure that nbd
client goes to reconnect loop.
7. On node1 run the following command
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
This will make the connect() call of qemu-io at node2 take a long time.
And you'll see that read command in qemu-io will hang for a long time,
more than 15 seconds specified by reconnect-delay parameter. It's the
bug.
8. Don't forget to drop iptables rule on node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Important note: Step [5] is necessary to reproduce _this_ bug. If we
miss step [5], the read command (step 6) will hang for a long time and
this commit doesn't help, because there will be not long connect() to
unreachable host, but long sendmsg() to unreachable host, which should
be fixed by enabling and adjusting keep-alive on the socket, which is a
thing for further patch set.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200903190301.367620-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-09-03 21:03:00 +02:00
|
|
|
s->state = NBD_CLIENT_CONNECTING_NOWAIT;
|
|
|
|
while (qemu_co_enter_next(&s->free_sema, NULL)) {
|
|
|
|
/* Resume all queued requests */
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
reconnect_delay_timer_del(s);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void reconnect_delay_timer_init(BDRVNBDState *s, uint64_t expire_time_ns)
|
|
|
|
{
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) != NBD_CLIENT_CONNECTING_WAIT) {
|
block/nbd: fix reconnect-delay
reconnect-delay has a design flaw: we handle it in the same loop where
we do connection attempt. So, reconnect-delay may be exceeded by
unpredictable time of connection attempt.
Let's instead use separate timer.
How to reproduce the bug:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. On node2 start qemu-io:
./build/qemu-io --image-opts \
driver=nbd,server.type=inet,server.host=192.168.100.5,server.port=10809,reconnect-delay=15
4. Type 'read 0 512' in qemu-io interface to check that connection
works
Be careful: you should make steps 5-7 in a short time, less than 15
seconds.
5. Kill nbd server on node1
6. Run 'read 0 512' in qemu-io interface again, to be sure that nbd
client goes to reconnect loop.
7. On node1 run the following command
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
This will make the connect() call of qemu-io at node2 take a long time.
And you'll see that read command in qemu-io will hang for a long time,
more than 15 seconds specified by reconnect-delay parameter. It's the
bug.
8. Don't forget to drop iptables rule on node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Important note: Step [5] is necessary to reproduce _this_ bug. If we
miss step [5], the read command (step 6) will hang for a long time and
this commit doesn't help, because there will be not long connect() to
unreachable host, but long sendmsg() to unreachable host, which should
be fixed by enabling and adjusting keep-alive on the socket, which is a
thing for further patch set.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200903190301.367620-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-09-03 21:03:00 +02:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
assert(!s->reconnect_delay_timer);
|
|
|
|
s->reconnect_delay_timer = aio_timer_new(bdrv_get_aio_context(s->bs),
|
|
|
|
QEMU_CLOCK_REALTIME,
|
|
|
|
SCALE_NS,
|
|
|
|
reconnect_delay_timer_cb, s);
|
|
|
|
timer_mod(s->reconnect_delay_timer, expire_time_ns);
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:19 +02:00
|
|
|
static void nbd_client_detach_aio_context(BlockDriverState *bs)
|
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
|
|
|
|
block/nbd: fix reconnect-delay
reconnect-delay has a design flaw: we handle it in the same loop where
we do connection attempt. So, reconnect-delay may be exceeded by
unpredictable time of connection attempt.
Let's instead use separate timer.
How to reproduce the bug:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. On node2 start qemu-io:
./build/qemu-io --image-opts \
driver=nbd,server.type=inet,server.host=192.168.100.5,server.port=10809,reconnect-delay=15
4. Type 'read 0 512' in qemu-io interface to check that connection
works
Be careful: you should make steps 5-7 in a short time, less than 15
seconds.
5. Kill nbd server on node1
6. Run 'read 0 512' in qemu-io interface again, to be sure that nbd
client goes to reconnect loop.
7. On node1 run the following command
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
This will make the connect() call of qemu-io at node2 take a long time.
And you'll see that read command in qemu-io will hang for a long time,
more than 15 seconds specified by reconnect-delay parameter. It's the
bug.
8. Don't forget to drop iptables rule on node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Important note: Step [5] is necessary to reproduce _this_ bug. If we
miss step [5], the read command (step 6) will hang for a long time and
this commit doesn't help, because there will be not long connect() to
unreachable host, but long sendmsg() to unreachable host, which should
be fixed by enabling and adjusting keep-alive on the socket, which is a
thing for further patch set.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200903190301.367620-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-09-03 21:03:00 +02:00
|
|
|
/* Timer is deleted in nbd_client_co_drain_begin() */
|
|
|
|
assert(!s->reconnect_delay_timer);
|
block/nbd: only detach existing iochannel from aio_context
When the reconnect in NBD client is in progress, the iochannel used for
NBD connection doesn't exist. Therefore an attempt to detach it from
the aio_context of the parent BlockDriverState results in a NULL pointer
dereference.
The problem is triggerable, in particular, when an outgoing migration is
about to finish, and stopping the dataplane tries to move the
BlockDriverState from the iothread aio_context to the main loop. If the
NBD connection is lost before this point, and the NBD client has entered
the reconnect procedure, QEMU crashes:
#0 qemu_aio_coroutine_enter (ctx=0x5618056c7580, co=0x0)
at /build/qemu-6MF7tq/qemu-5.0.1/util/qemu-coroutine.c:109
#1 0x00005618034b1b68 in nbd_client_attach_aio_context_bh (
opaque=0x561805ed4c00) at /build/qemu-6MF7tq/qemu-5.0.1/block/nbd.c:164
#2 0x000056180353116b in aio_wait_bh (opaque=0x7f60e1e63700)
at /build/qemu-6MF7tq/qemu-5.0.1/util/aio-wait.c:55
#3 0x0000561803530633 in aio_bh_call (bh=0x7f60d40a7e80)
at /build/qemu-6MF7tq/qemu-5.0.1/util/async.c:136
#4 aio_bh_poll (ctx=ctx@entry=0x5618056c7580)
at /build/qemu-6MF7tq/qemu-5.0.1/util/async.c:164
#5 0x0000561803533e5a in aio_poll (ctx=ctx@entry=0x5618056c7580,
blocking=blocking@entry=true)
at /build/qemu-6MF7tq/qemu-5.0.1/util/aio-posix.c:650
#6 0x000056180353128d in aio_wait_bh_oneshot (ctx=0x5618056c7580,
cb=<optimized out>, opaque=<optimized out>)
at /build/qemu-6MF7tq/qemu-5.0.1/util/aio-wait.c:71
#7 0x000056180345c50a in bdrv_attach_aio_context (new_context=0x5618056c7580,
bs=0x561805ed4c00) at /build/qemu-6MF7tq/qemu-5.0.1/block.c:6172
#8 bdrv_set_aio_context_ignore (bs=bs@entry=0x561805ed4c00,
new_context=new_context@entry=0x5618056c7580,
ignore=ignore@entry=0x7f60e1e63780)
at /build/qemu-6MF7tq/qemu-5.0.1/block.c:6237
#9 0x000056180345c969 in bdrv_child_try_set_aio_context (
bs=bs@entry=0x561805ed4c00, ctx=0x5618056c7580,
ignore_child=<optimized out>, errp=<optimized out>)
at /build/qemu-6MF7tq/qemu-5.0.1/block.c:6332
#10 0x00005618034957db in blk_do_set_aio_context (blk=0x56180695b3f0,
new_context=0x5618056c7580, update_root_node=update_root_node@entry=true,
errp=errp@entry=0x0)
at /build/qemu-6MF7tq/qemu-5.0.1/block/block-backend.c:1989
#11 0x00005618034980bd in blk_set_aio_context (blk=<optimized out>,
new_context=<optimized out>, errp=errp@entry=0x0)
at /build/qemu-6MF7tq/qemu-5.0.1/block/block-backend.c:2010
#12 0x0000561803197953 in virtio_blk_data_plane_stop (vdev=<optimized out>)
at /build/qemu-6MF7tq/qemu-5.0.1/hw/block/dataplane/virtio-blk.c:292
#13 0x00005618033d67bf in virtio_bus_stop_ioeventfd (bus=0x5618056d9f08)
at /build/qemu-6MF7tq/qemu-5.0.1/hw/virtio/virtio-bus.c:245
#14 0x00005618031c9b2e in virtio_vmstate_change (opaque=0x5618056d9f90,
running=0, state=<optimized out>)
at /build/qemu-6MF7tq/qemu-5.0.1/hw/virtio/virtio.c:3220
#15 0x0000561803208bfd in vm_state_notify (running=running@entry=0,
state=state@entry=RUN_STATE_FINISH_MIGRATE)
at /build/qemu-6MF7tq/qemu-5.0.1/softmmu/vl.c:1275
#16 0x0000561803155c02 in do_vm_stop (state=RUN_STATE_FINISH_MIGRATE,
send_stop=<optimized out>) at /build/qemu-6MF7tq/qemu-5.0.1/cpus.c:1032
#17 0x00005618033e3765 in migration_completion (s=0x5618056e6960)
at /build/qemu-6MF7tq/qemu-5.0.1/migration/migration.c:2914
#18 migration_iteration_run (s=0x5618056e6960)
at /build/qemu-6MF7tq/qemu-5.0.1/migration/migration.c:3275
#19 migration_thread (opaque=opaque@entry=0x5618056e6960)
at /build/qemu-6MF7tq/qemu-5.0.1/migration/migration.c:3439
#20 0x0000561803536ad6 in qemu_thread_start (args=<optimized out>)
at /build/qemu-6MF7tq/qemu-5.0.1/util/qemu-thread-posix.c:519
#21 0x00007f61085d06ba in start_thread ()
from /lib/x86_64-linux-gnu/libpthread.so.0
#22 0x00007f610830641d in sysctl () from /lib/x86_64-linux-gnu/libc.so.6
#23 0x0000000000000000 in ?? ()
Fix it by checking that the iochannel is non-null before trying to
detach it from the aio_context. If it is null, no detaching is needed,
and it will get reattached in the proper aio_context once the connection
is reestablished.
Signed-off-by: Roman Kagan <rvkagan@yandex-team.ru>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210129073859.683063-2-rvkagan@yandex-team.ru>
Signed-off-by: Eric Blake <eblake@redhat.com>
2021-01-29 08:38:57 +01:00
|
|
|
/*
|
|
|
|
* If reconnect is in progress we may have no ->ioc. It will be
|
|
|
|
* re-instantiated in the proper aio context once the connection is
|
|
|
|
* reestablished.
|
|
|
|
*/
|
|
|
|
if (s->ioc) {
|
|
|
|
qio_channel_detach_aio_context(QIO_CHANNEL(s->ioc));
|
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void nbd_client_attach_aio_context_bh(void *opaque)
|
|
|
|
{
|
|
|
|
BlockDriverState *bs = opaque;
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
|
block/nbd: only enter connection coroutine if it's present
When an NBD block driver state is moved from one aio_context to another
(e.g. when doing a drain in a migration thread),
nbd_client_attach_aio_context_bh is executed that enters the connection
coroutine.
However, the assumption that ->connection_co is always present here
appears incorrect: the connection may have encountered an error other
than -EIO in the underlying transport, and thus may have decided to quit
rather than keep trying to reconnect, and therefore it may have
terminated the connection coroutine. As a result an attempt to reassign
the client in this state (NBD_CLIENT_QUIT) to a different aio_context
leads to a null pointer dereference:
#0 qio_channel_detach_aio_context (ioc=0x0)
at /build/qemu-gYtjVn/qemu-5.0.1/io/channel.c:452
#1 0x0000562a242824b3 in bdrv_detach_aio_context (bs=0x562a268d6a00)
at /build/qemu-gYtjVn/qemu-5.0.1/block.c:6151
#2 bdrv_set_aio_context_ignore (bs=bs@entry=0x562a268d6a00,
new_context=new_context@entry=0x562a260c9580,
ignore=ignore@entry=0x7feeadc9b780)
at /build/qemu-gYtjVn/qemu-5.0.1/block.c:6230
#3 0x0000562a24282969 in bdrv_child_try_set_aio_context
(bs=bs@entry=0x562a268d6a00, ctx=0x562a260c9580,
ignore_child=<optimized out>, errp=<optimized out>)
at /build/qemu-gYtjVn/qemu-5.0.1/block.c:6332
#4 0x0000562a242bb7db in blk_do_set_aio_context (blk=0x562a2735d0d0,
new_context=0x562a260c9580,
update_root_node=update_root_node@entry=true, errp=errp@entry=0x0)
at /build/qemu-gYtjVn/qemu-5.0.1/block/block-backend.c:1989
#5 0x0000562a242be0bd in blk_set_aio_context (blk=<optimized out>,
new_context=<optimized out>, errp=errp@entry=0x0)
at /build/qemu-gYtjVn/qemu-5.0.1/block/block-backend.c:2010
#6 0x0000562a23fbd953 in virtio_blk_data_plane_stop (vdev=<optimized
out>)
at /build/qemu-gYtjVn/qemu-5.0.1/hw/block/dataplane/virtio-blk.c:292
#7 0x0000562a241fc7bf in virtio_bus_stop_ioeventfd (bus=0x562a260dbf08)
at /build/qemu-gYtjVn/qemu-5.0.1/hw/virtio/virtio-bus.c:245
#8 0x0000562a23fefb2e in virtio_vmstate_change (opaque=0x562a260dbf90,
running=0, state=<optimized out>)
at /build/qemu-gYtjVn/qemu-5.0.1/hw/virtio/virtio.c:3220
#9 0x0000562a2402ebfd in vm_state_notify (running=running@entry=0,
state=state@entry=RUN_STATE_FINISH_MIGRATE)
at /build/qemu-gYtjVn/qemu-5.0.1/softmmu/vl.c:1275
#10 0x0000562a23f7bc02 in do_vm_stop (state=RUN_STATE_FINISH_MIGRATE,
send_stop=<optimized out>)
at /build/qemu-gYtjVn/qemu-5.0.1/cpus.c:1032
#11 0x0000562a24209765 in migration_completion (s=0x562a260e83a0)
at /build/qemu-gYtjVn/qemu-5.0.1/migration/migration.c:2914
#12 migration_iteration_run (s=0x562a260e83a0)
at /build/qemu-gYtjVn/qemu-5.0.1/migration/migration.c:3275
#13 migration_thread (opaque=opaque@entry=0x562a260e83a0)
at /build/qemu-gYtjVn/qemu-5.0.1/migration/migration.c:3439
#14 0x0000562a2435ca96 in qemu_thread_start (args=<optimized out>)
at /build/qemu-gYtjVn/qemu-5.0.1/util/qemu-thread-posix.c:519
#15 0x00007feed31466ba in start_thread (arg=0x7feeadc9c700)
at pthread_create.c:333
#16 0x00007feed2e7c41d in __GI___sysctl (name=0x0, nlen=608471908,
oldval=0x562a2452b138, oldlenp=0x0, newval=0x562a2452c5e0
<__func__.28102>, newlen=0)
at ../sysdeps/unix/sysv/linux/sysctl.c:30
#17 0x0000000000000000 in ?? ()
Fix it by checking that the connection coroutine is non-null before
trying to enter it. If it is null, no entering is needed, as the
connection is probably going down anyway.
Signed-off-by: Roman Kagan <rvkagan@yandex-team.ru>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210129073859.683063-3-rvkagan@yandex-team.ru>
Signed-off-by: Eric Blake <eblake@redhat.com>
2021-01-29 08:38:58 +01:00
|
|
|
if (s->connection_co) {
|
|
|
|
/*
|
|
|
|
* The node is still drained, so we know the coroutine has yielded in
|
|
|
|
* nbd_read_eof(), the only place where bs->in_flight can reach 0, or
|
|
|
|
* it is entered for the first time. Both places are safe for entering
|
|
|
|
* the coroutine.
|
|
|
|
*/
|
|
|
|
qemu_aio_coroutine_enter(bs->aio_context, s->connection_co);
|
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
bdrv_dec_in_flight(bs);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nbd_client_attach_aio_context(BlockDriverState *bs,
|
|
|
|
AioContext *new_context)
|
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
/*
|
|
|
|
* s->connection_co is either yielded from nbd_receive_reply or from
|
|
|
|
* nbd_co_reconnect_loop()
|
|
|
|
*/
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTED) {
|
2019-10-09 10:41:57 +02:00
|
|
|
qio_channel_attach_aio_context(QIO_CHANNEL(s->ioc), new_context);
|
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
bdrv_inc_in_flight(bs);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Need to wait here for the BH to run because the BH must run while the
|
|
|
|
* node is still drained.
|
|
|
|
*/
|
|
|
|
aio_wait_bh_oneshot(new_context, nbd_client_attach_aio_context_bh, bs);
|
|
|
|
}
|
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
static void coroutine_fn nbd_client_co_drain_begin(BlockDriverState *bs)
|
|
|
|
{
|
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
s->drained = true;
|
2021-05-17 12:05:47 +02:00
|
|
|
qemu_co_sleep_wake(&s->reconnect_sleep);
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
|
2021-06-10 12:07:37 +02:00
|
|
|
nbd_co_establish_connection_cancel(bs);
|
block/nbd: fix drain dead-lock because of nbd reconnect-delay
We pause reconnect process during drained section. So, if we have some
requests, waiting for reconnect we should cancel them, otherwise they
deadlock the drained section.
How to reproduce:
1. Create an image:
qemu-img create -f qcow2 xx 100M
2. Start NBD server:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./build/x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive \
driver=nbd,server.type=inet,server.host=192.168.100.5,server.port=10809,reconnect-delay=60 \
-vnc :0 -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, kill the nbd server, and run dd in the guest again:
dd if=/dev/sdb of=/dev/null bs=1M count=10
Now Qemu is trying to reconnect, and dd-generated requests are waiting
for the connection (they will wait up to 60 seconds (see
reconnect-delay option above) and than fail). But suddenly, vm may
totally hang in the deadlock. You may need to increase reconnect-delay
period to catch the dead-lock.
VM doesn't respond because drain dead-lock happens in cpu thread with
global mutex taken. That's not good thing by itself and is not fixed
by this commit (true way is using iothreads). Still this commit fixes
drain dead-lock itself.
Note: probably, we can instead continue to reconnect during drained
section. To achieve this, we may move negotiation to the connect thread
to make it independent of bs aio context. But expanding drained section
doesn't seem good anyway. So, let's now fix the bug the simplest way.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200903190301.367620-2-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-09-03 21:02:58 +02:00
|
|
|
|
block/nbd: fix reconnect-delay
reconnect-delay has a design flaw: we handle it in the same loop where
we do connection attempt. So, reconnect-delay may be exceeded by
unpredictable time of connection attempt.
Let's instead use separate timer.
How to reproduce the bug:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. On node2 start qemu-io:
./build/qemu-io --image-opts \
driver=nbd,server.type=inet,server.host=192.168.100.5,server.port=10809,reconnect-delay=15
4. Type 'read 0 512' in qemu-io interface to check that connection
works
Be careful: you should make steps 5-7 in a short time, less than 15
seconds.
5. Kill nbd server on node1
6. Run 'read 0 512' in qemu-io interface again, to be sure that nbd
client goes to reconnect loop.
7. On node1 run the following command
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
This will make the connect() call of qemu-io at node2 take a long time.
And you'll see that read command in qemu-io will hang for a long time,
more than 15 seconds specified by reconnect-delay parameter. It's the
bug.
8. Don't forget to drop iptables rule on node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Important note: Step [5] is necessary to reproduce _this_ bug. If we
miss step [5], the read command (step 6) will hang for a long time and
this commit doesn't help, because there will be not long connect() to
unreachable host, but long sendmsg() to unreachable host, which should
be fixed by enabling and adjusting keep-alive on the socket, which is a
thing for further patch set.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200903190301.367620-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-09-03 21:03:00 +02:00
|
|
|
reconnect_delay_timer_del(s);
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT) {
|
block/nbd: fix drain dead-lock because of nbd reconnect-delay
We pause reconnect process during drained section. So, if we have some
requests, waiting for reconnect we should cancel them, otherwise they
deadlock the drained section.
How to reproduce:
1. Create an image:
qemu-img create -f qcow2 xx 100M
2. Start NBD server:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./build/x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive \
driver=nbd,server.type=inet,server.host=192.168.100.5,server.port=10809,reconnect-delay=60 \
-vnc :0 -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, kill the nbd server, and run dd in the guest again:
dd if=/dev/sdb of=/dev/null bs=1M count=10
Now Qemu is trying to reconnect, and dd-generated requests are waiting
for the connection (they will wait up to 60 seconds (see
reconnect-delay option above) and than fail). But suddenly, vm may
totally hang in the deadlock. You may need to increase reconnect-delay
period to catch the dead-lock.
VM doesn't respond because drain dead-lock happens in cpu thread with
global mutex taken. That's not good thing by itself and is not fixed
by this commit (true way is using iothreads). Still this commit fixes
drain dead-lock itself.
Note: probably, we can instead continue to reconnect during drained
section. To achieve this, we may move negotiation to the connect thread
to make it independent of bs aio context. But expanding drained section
doesn't seem good anyway. So, let's now fix the bug the simplest way.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200903190301.367620-2-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-09-03 21:02:58 +02:00
|
|
|
s->state = NBD_CLIENT_CONNECTING_NOWAIT;
|
|
|
|
qemu_co_queue_restart_all(&s->free_sema);
|
|
|
|
}
|
2019-10-09 10:41:57 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void coroutine_fn nbd_client_co_drain_end(BlockDriverState *bs)
|
2019-06-11 12:27:19 +02:00
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
s->drained = false;
|
|
|
|
if (s->wait_drained_end) {
|
|
|
|
s->wait_drained_end = false;
|
|
|
|
aio_co_wake(s->connection_co);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
static void nbd_teardown_connection(BlockDriverState *bs)
|
|
|
|
{
|
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
|
|
|
|
2020-07-27 20:47:49 +02:00
|
|
|
if (s->ioc) {
|
2019-10-09 10:41:57 +02:00
|
|
|
/* finish any pending coroutines */
|
|
|
|
qio_channel_shutdown(s->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL);
|
2020-07-27 20:47:49 +02:00
|
|
|
} else if (s->sioc) {
|
|
|
|
/* abort negotiation */
|
|
|
|
qio_channel_shutdown(QIO_CHANNEL(s->sioc), QIO_CHANNEL_SHUTDOWN_BOTH,
|
|
|
|
NULL);
|
2019-10-09 10:41:57 +02:00
|
|
|
}
|
2020-07-27 20:47:49 +02:00
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
s->state = NBD_CLIENT_QUIT;
|
|
|
|
if (s->connection_co) {
|
2021-05-17 12:05:47 +02:00
|
|
|
qemu_co_sleep_wake(&s->reconnect_sleep);
|
2021-06-10 12:07:37 +02:00
|
|
|
nbd_co_establish_connection_cancel(bs);
|
2019-10-09 10:41:57 +02:00
|
|
|
}
|
2020-01-22 17:45:28 +01:00
|
|
|
if (qemu_in_coroutine()) {
|
|
|
|
s->teardown_co = qemu_coroutine_self();
|
|
|
|
/* connection_co resumes us when it terminates */
|
|
|
|
qemu_coroutine_yield();
|
|
|
|
s->teardown_co = NULL;
|
|
|
|
} else {
|
|
|
|
BDRV_POLL_WHILE(bs, s->connection_co);
|
|
|
|
}
|
|
|
|
assert(!s->connection_co);
|
2019-10-09 10:41:57 +02:00
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
static bool nbd_client_connecting(BDRVNBDState *s)
|
|
|
|
{
|
2020-12-28 16:08:44 +01:00
|
|
|
NBDClientState state = qatomic_load_acquire(&s->state);
|
|
|
|
return state == NBD_CLIENT_CONNECTING_WAIT ||
|
|
|
|
state == NBD_CLIENT_CONNECTING_NOWAIT;
|
2019-10-09 10:41:57 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static bool nbd_client_connecting_wait(BDRVNBDState *s)
|
|
|
|
{
|
2020-12-28 16:08:44 +01:00
|
|
|
return qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT;
|
2019-10-09 10:41:57 +02:00
|
|
|
}
|
|
|
|
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
static void nbd_init_connect_thread(BDRVNBDState *s)
|
|
|
|
{
|
|
|
|
s->connect_thread = g_new(NBDConnectThread, 1);
|
|
|
|
|
|
|
|
*s->connect_thread = (NBDConnectThread) {
|
|
|
|
.saddr = QAPI_CLONE(SocketAddress, s->saddr),
|
|
|
|
};
|
|
|
|
|
|
|
|
qemu_mutex_init(&s->connect_thread->mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nbd_free_connect_thread(NBDConnectThread *thr)
|
|
|
|
{
|
|
|
|
if (thr->sioc) {
|
|
|
|
qio_channel_close(QIO_CHANNEL(thr->sioc), NULL);
|
2021-06-10 12:07:32 +02:00
|
|
|
object_unref(OBJECT(thr->sioc));
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
}
|
|
|
|
error_free(thr->err);
|
|
|
|
qapi_free_SocketAddress(thr->saddr);
|
|
|
|
g_free(thr);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void *connect_thread_func(void *opaque)
|
|
|
|
{
|
|
|
|
NBDConnectThread *thr = opaque;
|
|
|
|
int ret;
|
2021-06-10 12:07:41 +02:00
|
|
|
bool do_free;
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
|
|
|
|
thr->sioc = qio_channel_socket_new();
|
|
|
|
|
|
|
|
error_free(thr->err);
|
|
|
|
thr->err = NULL;
|
|
|
|
ret = qio_channel_socket_connect_sync(thr->sioc, thr->saddr, &thr->err);
|
|
|
|
if (ret < 0) {
|
|
|
|
object_unref(OBJECT(thr->sioc));
|
|
|
|
thr->sioc = NULL;
|
|
|
|
}
|
|
|
|
|
2021-06-10 12:07:34 +02:00
|
|
|
qio_channel_set_delay(QIO_CHANNEL(thr->sioc), false);
|
|
|
|
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
qemu_mutex_lock(&thr->mutex);
|
|
|
|
|
2021-06-10 12:07:41 +02:00
|
|
|
assert(thr->running);
|
|
|
|
thr->running = false;
|
|
|
|
if (thr->wait_co) {
|
|
|
|
aio_co_wake(thr->wait_co);
|
|
|
|
thr->wait_co = NULL;
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
}
|
2021-06-10 12:07:41 +02:00
|
|
|
do_free = thr->detached;
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
|
|
|
|
qemu_mutex_unlock(&thr->mutex);
|
|
|
|
|
|
|
|
if (do_free) {
|
|
|
|
nbd_free_connect_thread(thr);
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2021-06-10 12:07:42 +02:00
|
|
|
/*
|
|
|
|
* Get a new connection in context of @thr:
|
|
|
|
* if the thread is running, wait for completion
|
|
|
|
* if the thread already succeeded in the background, and user didn't get the
|
|
|
|
* result, just return it now
|
|
|
|
* otherwise the thread is not running, so start a thread and wait for
|
|
|
|
* completion
|
|
|
|
*/
|
|
|
|
static coroutine_fn QIOChannelSocket *
|
|
|
|
nbd_co_establish_connection(NBDConnectThread *thr, Error **errp)
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
{
|
2021-06-10 12:07:42 +02:00
|
|
|
QIOChannelSocket *sioc = NULL;
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
QemuThread thread;
|
2021-06-10 12:07:41 +02:00
|
|
|
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
qemu_mutex_lock(&thr->mutex);
|
|
|
|
|
2021-06-10 12:07:42 +02:00
|
|
|
/*
|
|
|
|
* Don't call nbd_co_establish_connection() in several coroutines in
|
|
|
|
* parallel. Only one call at once is supported.
|
|
|
|
*/
|
|
|
|
assert(!thr->wait_co);
|
|
|
|
|
2021-06-10 12:07:41 +02:00
|
|
|
if (!thr->running) {
|
|
|
|
if (thr->sioc) {
|
|
|
|
/* Previous attempt finally succeeded in background */
|
2021-06-10 12:07:42 +02:00
|
|
|
sioc = g_steal_pointer(&thr->sioc);
|
|
|
|
qemu_mutex_unlock(&thr->mutex);
|
|
|
|
|
|
|
|
return sioc;
|
2021-06-10 12:07:41 +02:00
|
|
|
}
|
2021-06-10 12:07:42 +02:00
|
|
|
|
2021-06-10 12:07:41 +02:00
|
|
|
thr->running = true;
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
error_free(thr->err);
|
|
|
|
thr->err = NULL;
|
|
|
|
qemu_thread_create(&thread, "nbd-connect",
|
|
|
|
connect_thread_func, thr, QEMU_THREAD_DETACHED);
|
|
|
|
}
|
|
|
|
|
block/nbd: simplify waking of nbd_co_establish_connection()
Instead of managing connect_bh, bh_ctx, and wait_connect fields, we
can use a single link to the waiting coroutine with proper mutex
protection.
So new logic is:
nbd_co_establish_connection() sets wait_co under the mutex, releases
the mutex, then yield()s. Note that wait_co may be scheduled by the
thread immediately after unlocking the mutex. Still, the main thread
(or iothread) will not reach the code for entering the coroutine until
the yield(), so we are safe.
connect_thread_func() and nbd_co_establish_connection_cancel() do
the following to handle wait_co:
Under the mutex, if thr->wait_co is not NULL, make it NULL and
schedule it. This way, we avoid scheduling the coroutine twice.
Still scheduling is a bit different:
In connect_thread_func() we can just call aio_co_wake under mutex,
after commit
[async: the main AioContext is only "current" if under the BQL]
we are sure that aio_co_wake() will not try to acquire the aio context
and do qemu_aio_coroutine_enter() but simply schedule the coroutine by
aio_co_schedule().
nbd_co_establish_connection_cancel() will be called from non-coroutine
context in further patch and will be able to go through
qemu_aio_coroutine_enter() path of aio_co_wake(). So keep current
behavior of waking the coroutine after the critical section.
Also, this commit reduces the dependence of
nbd_co_establish_connection() on the internals of bs (we now use a
generic pointer to the coroutine, instead of direct use of
s->connection_co). This is a step towards splitting the connection
API out of nbd.c.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210610100802.5888-11-vsementsov@virtuozzo.com>
Reviewied-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2021-06-10 12:07:40 +02:00
|
|
|
thr->wait_co = qemu_coroutine_self();
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
|
|
|
|
qemu_mutex_unlock(&thr->mutex);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We are going to wait for connect-thread finish, but
|
2021-06-10 12:07:42 +02:00
|
|
|
* nbd_co_establish_connection_cancel() can interrupt.
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
*/
|
|
|
|
qemu_coroutine_yield();
|
|
|
|
|
|
|
|
qemu_mutex_lock(&thr->mutex);
|
|
|
|
|
2021-06-10 12:07:41 +02:00
|
|
|
if (thr->running) {
|
|
|
|
/*
|
|
|
|
* The connection attempt was canceled and the coroutine resumed
|
|
|
|
* before the connection thread finished its job. Report the
|
|
|
|
* attempt as failed, but leave the connection thread running,
|
|
|
|
* to reuse it for the next connection attempt.
|
|
|
|
*/
|
|
|
|
error_setg(errp, "Connection attempt cancelled by other operation");
|
|
|
|
} else {
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
error_propagate(errp, thr->err);
|
|
|
|
thr->err = NULL;
|
2021-06-10 12:07:42 +02:00
|
|
|
sioc = g_steal_pointer(&thr->sioc);
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
qemu_mutex_unlock(&thr->mutex);
|
|
|
|
|
2021-06-10 12:07:42 +02:00
|
|
|
return sioc;
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nbd_co_establish_connection_cancel
|
|
|
|
* Cancel nbd_co_establish_connection asynchronously: it will finish soon, to
|
|
|
|
* allow drained section to begin.
|
|
|
|
*/
|
2021-06-10 12:07:37 +02:00
|
|
|
static void nbd_co_establish_connection_cancel(BlockDriverState *bs)
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
{
|
|
|
|
BDRVNBDState *s = bs->opaque;
|
|
|
|
NBDConnectThread *thr = s->connect_thread;
|
2021-06-10 12:07:41 +02:00
|
|
|
Coroutine *wait_co;
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
|
|
|
|
qemu_mutex_lock(&thr->mutex);
|
|
|
|
|
2021-06-10 12:07:41 +02:00
|
|
|
wait_co = g_steal_pointer(&thr->wait_co);
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
|
|
|
|
qemu_mutex_unlock(&thr->mutex);
|
|
|
|
|
block/nbd: simplify waking of nbd_co_establish_connection()
Instead of managing connect_bh, bh_ctx, and wait_connect fields, we
can use a single link to the waiting coroutine with proper mutex
protection.
So new logic is:
nbd_co_establish_connection() sets wait_co under the mutex, releases
the mutex, then yield()s. Note that wait_co may be scheduled by the
thread immediately after unlocking the mutex. Still, the main thread
(or iothread) will not reach the code for entering the coroutine until
the yield(), so we are safe.
connect_thread_func() and nbd_co_establish_connection_cancel() do
the following to handle wait_co:
Under the mutex, if thr->wait_co is not NULL, make it NULL and
schedule it. This way, we avoid scheduling the coroutine twice.
Still scheduling is a bit different:
In connect_thread_func() we can just call aio_co_wake under mutex,
after commit
[async: the main AioContext is only "current" if under the BQL]
we are sure that aio_co_wake() will not try to acquire the aio context
and do qemu_aio_coroutine_enter() but simply schedule the coroutine by
aio_co_schedule().
nbd_co_establish_connection_cancel() will be called from non-coroutine
context in further patch and will be able to go through
qemu_aio_coroutine_enter() path of aio_co_wake(). So keep current
behavior of waking the coroutine after the critical section.
Also, this commit reduces the dependence of
nbd_co_establish_connection() on the internals of bs (we now use a
generic pointer to the coroutine, instead of direct use of
s->connection_co). This is a step towards splitting the connection
API out of nbd.c.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210610100802.5888-11-vsementsov@virtuozzo.com>
Reviewied-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2021-06-10 12:07:40 +02:00
|
|
|
if (wait_co) {
|
|
|
|
aio_co_wake(wait_co);
|
block/nbd: use non-blocking connect: fix vm hang on connect()
This makes nbd's connection_co yield during reconnects, so that
reconnect doesn't block the main thread. This is very important in
case of an unavailable nbd server host: connect() call may take a long
time, blocking the main thread (and due to reconnect, it will hang
again and again with small gaps of working time during pauses between
connection attempts).
Realization notes:
- We don't want to implement non-blocking connect() over non-blocking
socket, because getaddrinfo() doesn't have portable non-blocking
realization anyway, so let's just use a thread for both getaddrinfo()
and connect().
- We can't use qio_channel_socket_connect_async (which behaves
similarly and starts a thread to execute connect() call), as it's relying
on someone iterating main loop (g_main_loop_run() or something like
this), which is not always the case.
- We can't use thread_pool_submit_co API, as thread pool waits for all
threads to finish (but we don't want to wait for blocking reconnect
attempt on shutdown.
So, we just create the thread by hand. Some additional difficulties
are:
- We want our connect to avoid blocking drained sections and aio context
switches. To achieve this, we make it possible to "cancel" synchronous
wait for the connect (which is a coroutine yield actually), still,
the thread continues in background, and if successful, its result may be
reused on next reconnect attempt.
- We don't want to wait for reconnect on shutdown, so there is
CONNECT_THREAD_RUNNING_DETACHED thread state, which means that the block
layer is no longer interested in a result, and thread should close new
connected socket on finish and free the state.
How to reproduce the bug, fixed with this commit:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. Start vm with second nbd disk on node2, like this:
./x86_64-softmmu/qemu-system-x86_64 -nodefaults -drive \
file=/work/images/cent7.qcow2 -drive file=nbd+tcp://192.168.100.2 \
-vnc :0 -qmp stdio -m 2G -enable-kvm -vga std
4. Access the vm through vnc (or some other way?), and check that NBD
drive works:
dd if=/dev/sdb of=/dev/null bs=1M count=10
- the command should succeed.
5. Now, let's trigger nbd-reconnect loop in Qemu process. For this:
5.1 Kill NBD server on node1
5.2 run "dd if=/dev/sdb of=/dev/null bs=1M count=10" in the guest
again. The command should fail and a lot of error messages about
failing disk may appear as well.
Now NBD client driver in Qemu tries to reconnect.
Still, VM works well.
6. Make node1 unavailable on NBD port, so connect() from node2 will
last for a long time:
On node1 (Note, that 10809 is just a default NBD port):
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
After some time the guest hangs, and you may check in gdb that Qemu
hangs in connect() call, issued from the main thread. This is the
BUG.
7. Don't forget to drop iptables rule from your node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200812145237.4396-1-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: minor wording and formatting tweaks]
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-08-12 16:52:37 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
static coroutine_fn void nbd_reconnect_attempt(BDRVNBDState *s)
|
|
|
|
{
|
2020-07-27 20:47:47 +02:00
|
|
|
int ret;
|
2019-10-09 10:41:57 +02:00
|
|
|
|
|
|
|
if (!nbd_client_connecting(s)) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Wait for completion of all in-flight requests */
|
|
|
|
|
|
|
|
qemu_co_mutex_lock(&s->send_mutex);
|
|
|
|
|
|
|
|
while (s->in_flight > 0) {
|
|
|
|
qemu_co_mutex_unlock(&s->send_mutex);
|
|
|
|
nbd_recv_coroutines_wake_all(s);
|
|
|
|
s->wait_in_flight = true;
|
|
|
|
qemu_coroutine_yield();
|
|
|
|
s->wait_in_flight = false;
|
|
|
|
qemu_co_mutex_lock(&s->send_mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
qemu_co_mutex_unlock(&s->send_mutex);
|
|
|
|
|
|
|
|
if (!nbd_client_connecting(s)) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Now we are sure that nobody is accessing the channel, and no one will
|
|
|
|
* try until we set the state to CONNECTED.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* Finalize previous connection if any */
|
|
|
|
if (s->ioc) {
|
2020-09-03 21:02:59 +02:00
|
|
|
qio_channel_detach_aio_context(QIO_CHANNEL(s->ioc));
|
2020-12-28 16:08:44 +01:00
|
|
|
yank_unregister_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name),
|
|
|
|
nbd_yank, s->bs);
|
2019-10-09 10:41:57 +02:00
|
|
|
object_unref(OBJECT(s->sioc));
|
|
|
|
s->sioc = NULL;
|
|
|
|
object_unref(OBJECT(s->ioc));
|
|
|
|
s->ioc = NULL;
|
|
|
|
}
|
|
|
|
|
2021-06-10 12:07:42 +02:00
|
|
|
s->sioc = nbd_co_establish_connection(s->connect_thread, NULL);
|
|
|
|
if (!s->sioc) {
|
2020-07-27 20:47:47 +02:00
|
|
|
ret = -ECONNREFUSED;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2021-06-10 12:07:42 +02:00
|
|
|
yank_register_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name), nbd_yank,
|
|
|
|
s->bs);
|
|
|
|
|
block/nbd: allow drain during reconnect attempt
It should be safe to reenter qio_channel_yield() on io/channel read/write
path, so it's safe to reduce in_flight and allow attaching new aio
context. And no problem to allow drain itself: connection attempt is
not a guest request. Moreover, if remote server is down, we can hang
in negotiation, blocking drain section and provoking a dead lock.
How to reproduce the dead lock:
1. Create nbd-fault-injector.conf with the following contents:
[inject-error "mega1"]
event=data
io=readwrite
when=before
2. In one terminal run nbd-fault-injector in a loop, like this:
n=1; while true; do
echo $n; ((n++));
./nbd-fault-injector.py 127.0.0.1:10000 nbd-fault-injector.conf;
done
3. In another terminal run qemu-io in a loop, like this:
n=1; while true; do
echo $n; ((n++));
./qemu-io -c 'read 0 512' nbd://127.0.0.1:10000;
done
After some time, qemu-io will hang trying to drain, for example, like
this:
#3 aio_poll (ctx=0x55f006bdd890, blocking=true) at
util/aio-posix.c:600
#4 bdrv_do_drained_begin (bs=0x55f006bea710, recursive=false,
parent=0x0, ignore_bds_parents=false, poll=true) at block/io.c:427
#5 bdrv_drained_begin (bs=0x55f006bea710) at block/io.c:433
#6 blk_drain (blk=0x55f006befc80) at block/block-backend.c:1710
#7 blk_unref (blk=0x55f006befc80) at block/block-backend.c:498
#8 bdrv_open_inherit (filename=0x7fffba1563bc
"nbd+tcp://127.0.0.1:10000", reference=0x0, options=0x55f006be86d0,
flags=24578, parent=0x0, child_class=0x0, child_role=0,
errp=0x7fffba154620) at block.c:3491
#9 bdrv_open (filename=0x7fffba1563bc "nbd+tcp://127.0.0.1:10000",
reference=0x0, options=0x0, flags=16386, errp=0x7fffba154620) at
block.c:3513
#10 blk_new_open (filename=0x7fffba1563bc "nbd+tcp://127.0.0.1:10000",
reference=0x0, options=0x0, flags=16386, errp=0x7fffba154620) at
block/block-backend.c:421
And connection_co stack like this:
#0 qemu_coroutine_switch (from_=0x55f006bf2650, to_=0x7fe96e07d918,
action=COROUTINE_YIELD) at util/coroutine-ucontext.c:302
#1 qemu_coroutine_yield () at util/qemu-coroutine.c:193
#2 qio_channel_yield (ioc=0x55f006bb3c20, condition=G_IO_IN) at
io/channel.c:472
#3 qio_channel_readv_all_eof (ioc=0x55f006bb3c20, iov=0x7fe96d729bf0,
niov=1, errp=0x7fe96d729eb0) at io/channel.c:110
#4 qio_channel_readv_all (ioc=0x55f006bb3c20, iov=0x7fe96d729bf0,
niov=1, errp=0x7fe96d729eb0) at io/channel.c:143
#5 qio_channel_read_all (ioc=0x55f006bb3c20, buf=0x7fe96d729d28
"\300.\366\004\360U", buflen=8, errp=0x7fe96d729eb0) at
io/channel.c:247
#6 nbd_read (ioc=0x55f006bb3c20, buffer=0x7fe96d729d28, size=8,
desc=0x55f004f69644 "initial magic", errp=0x7fe96d729eb0) at
/work/src/qemu/master/include/block/nbd.h:365
#7 nbd_read64 (ioc=0x55f006bb3c20, val=0x7fe96d729d28,
desc=0x55f004f69644 "initial magic", errp=0x7fe96d729eb0) at
/work/src/qemu/master/include/block/nbd.h:391
#8 nbd_start_negotiate (aio_context=0x55f006bdd890,
ioc=0x55f006bb3c20, tlscreds=0x0, hostname=0x0,
outioc=0x55f006bf19f8, structured_reply=true,
zeroes=0x7fe96d729dca, errp=0x7fe96d729eb0) at nbd/client.c:904
#9 nbd_receive_negotiate (aio_context=0x55f006bdd890,
ioc=0x55f006bb3c20, tlscreds=0x0, hostname=0x0,
outioc=0x55f006bf19f8, info=0x55f006bf1a00, errp=0x7fe96d729eb0) at
nbd/client.c:1032
#10 nbd_client_connect (bs=0x55f006bea710, errp=0x7fe96d729eb0) at
block/nbd.c:1460
#11 nbd_reconnect_attempt (s=0x55f006bf19f0) at block/nbd.c:287
#12 nbd_co_reconnect_loop (s=0x55f006bf19f0) at block/nbd.c:309
#13 nbd_connection_entry (opaque=0x55f006bf19f0) at block/nbd.c:360
#14 coroutine_trampoline (i0=113190480, i1=22000) at
util/coroutine-ucontext.c:173
Note, that the hang may be
triggered by another bug, so the whole case is fixed only together with
commit "block/nbd: on shutdown terminate connection attempt".
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200727184751.15704-3-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-07-27 20:47:48 +02:00
|
|
|
bdrv_dec_in_flight(s->bs);
|
|
|
|
|
2021-06-10 12:07:39 +02:00
|
|
|
ret = nbd_client_handshake(s->bs, NULL);
|
2020-07-27 20:47:47 +02:00
|
|
|
|
block/nbd: allow drain during reconnect attempt
It should be safe to reenter qio_channel_yield() on io/channel read/write
path, so it's safe to reduce in_flight and allow attaching new aio
context. And no problem to allow drain itself: connection attempt is
not a guest request. Moreover, if remote server is down, we can hang
in negotiation, blocking drain section and provoking a dead lock.
How to reproduce the dead lock:
1. Create nbd-fault-injector.conf with the following contents:
[inject-error "mega1"]
event=data
io=readwrite
when=before
2. In one terminal run nbd-fault-injector in a loop, like this:
n=1; while true; do
echo $n; ((n++));
./nbd-fault-injector.py 127.0.0.1:10000 nbd-fault-injector.conf;
done
3. In another terminal run qemu-io in a loop, like this:
n=1; while true; do
echo $n; ((n++));
./qemu-io -c 'read 0 512' nbd://127.0.0.1:10000;
done
After some time, qemu-io will hang trying to drain, for example, like
this:
#3 aio_poll (ctx=0x55f006bdd890, blocking=true) at
util/aio-posix.c:600
#4 bdrv_do_drained_begin (bs=0x55f006bea710, recursive=false,
parent=0x0, ignore_bds_parents=false, poll=true) at block/io.c:427
#5 bdrv_drained_begin (bs=0x55f006bea710) at block/io.c:433
#6 blk_drain (blk=0x55f006befc80) at block/block-backend.c:1710
#7 blk_unref (blk=0x55f006befc80) at block/block-backend.c:498
#8 bdrv_open_inherit (filename=0x7fffba1563bc
"nbd+tcp://127.0.0.1:10000", reference=0x0, options=0x55f006be86d0,
flags=24578, parent=0x0, child_class=0x0, child_role=0,
errp=0x7fffba154620) at block.c:3491
#9 bdrv_open (filename=0x7fffba1563bc "nbd+tcp://127.0.0.1:10000",
reference=0x0, options=0x0, flags=16386, errp=0x7fffba154620) at
block.c:3513
#10 blk_new_open (filename=0x7fffba1563bc "nbd+tcp://127.0.0.1:10000",
reference=0x0, options=0x0, flags=16386, errp=0x7fffba154620) at
block/block-backend.c:421
And connection_co stack like this:
#0 qemu_coroutine_switch (from_=0x55f006bf2650, to_=0x7fe96e07d918,
action=COROUTINE_YIELD) at util/coroutine-ucontext.c:302
#1 qemu_coroutine_yield () at util/qemu-coroutine.c:193
#2 qio_channel_yield (ioc=0x55f006bb3c20, condition=G_IO_IN) at
io/channel.c:472
#3 qio_channel_readv_all_eof (ioc=0x55f006bb3c20, iov=0x7fe96d729bf0,
niov=1, errp=0x7fe96d729eb0) at io/channel.c:110
#4 qio_channel_readv_all (ioc=0x55f006bb3c20, iov=0x7fe96d729bf0,
niov=1, errp=0x7fe96d729eb0) at io/channel.c:143
#5 qio_channel_read_all (ioc=0x55f006bb3c20, buf=0x7fe96d729d28
"\300.\366\004\360U", buflen=8, errp=0x7fe96d729eb0) at
io/channel.c:247
#6 nbd_read (ioc=0x55f006bb3c20, buffer=0x7fe96d729d28, size=8,
desc=0x55f004f69644 "initial magic", errp=0x7fe96d729eb0) at
/work/src/qemu/master/include/block/nbd.h:365
#7 nbd_read64 (ioc=0x55f006bb3c20, val=0x7fe96d729d28,
desc=0x55f004f69644 "initial magic", errp=0x7fe96d729eb0) at
/work/src/qemu/master/include/block/nbd.h:391
#8 nbd_start_negotiate (aio_context=0x55f006bdd890,
ioc=0x55f006bb3c20, tlscreds=0x0, hostname=0x0,
outioc=0x55f006bf19f8, structured_reply=true,
zeroes=0x7fe96d729dca, errp=0x7fe96d729eb0) at nbd/client.c:904
#9 nbd_receive_negotiate (aio_context=0x55f006bdd890,
ioc=0x55f006bb3c20, tlscreds=0x0, hostname=0x0,
outioc=0x55f006bf19f8, info=0x55f006bf1a00, errp=0x7fe96d729eb0) at
nbd/client.c:1032
#10 nbd_client_connect (bs=0x55f006bea710, errp=0x7fe96d729eb0) at
block/nbd.c:1460
#11 nbd_reconnect_attempt (s=0x55f006bf19f0) at block/nbd.c:287
#12 nbd_co_reconnect_loop (s=0x55f006bf19f0) at block/nbd.c:309
#13 nbd_connection_entry (opaque=0x55f006bf19f0) at block/nbd.c:360
#14 coroutine_trampoline (i0=113190480, i1=22000) at
util/coroutine-ucontext.c:173
Note, that the hang may be
triggered by another bug, so the whole case is fixed only together with
commit "block/nbd: on shutdown terminate connection attempt".
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200727184751.15704-3-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-07-27 20:47:48 +02:00
|
|
|
if (s->drained) {
|
|
|
|
s->wait_drained_end = true;
|
|
|
|
while (s->drained) {
|
|
|
|
/*
|
|
|
|
* We may be entered once from nbd_client_attach_aio_context_bh
|
|
|
|
* and then from nbd_client_co_drain_end. So here is a loop.
|
|
|
|
*/
|
|
|
|
qemu_coroutine_yield();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
bdrv_inc_in_flight(s->bs);
|
|
|
|
|
2020-07-27 20:47:47 +02:00
|
|
|
out:
|
|
|
|
if (ret >= 0) {
|
|
|
|
/* successfully connected */
|
|
|
|
s->state = NBD_CLIENT_CONNECTED;
|
|
|
|
qemu_co_queue_restart_all(&s->free_sema);
|
2019-10-09 10:41:57 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static coroutine_fn void nbd_co_reconnect_loop(BDRVNBDState *s)
|
|
|
|
{
|
|
|
|
uint64_t timeout = 1 * NANOSECONDS_PER_SECOND;
|
|
|
|
uint64_t max_timeout = 16 * NANOSECONDS_PER_SECOND;
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTING_WAIT) {
|
block/nbd: fix reconnect-delay
reconnect-delay has a design flaw: we handle it in the same loop where
we do connection attempt. So, reconnect-delay may be exceeded by
unpredictable time of connection attempt.
Let's instead use separate timer.
How to reproduce the bug:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. On node2 start qemu-io:
./build/qemu-io --image-opts \
driver=nbd,server.type=inet,server.host=192.168.100.5,server.port=10809,reconnect-delay=15
4. Type 'read 0 512' in qemu-io interface to check that connection
works
Be careful: you should make steps 5-7 in a short time, less than 15
seconds.
5. Kill nbd server on node1
6. Run 'read 0 512' in qemu-io interface again, to be sure that nbd
client goes to reconnect loop.
7. On node1 run the following command
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
This will make the connect() call of qemu-io at node2 take a long time.
And you'll see that read command in qemu-io will hang for a long time,
more than 15 seconds specified by reconnect-delay parameter. It's the
bug.
8. Don't forget to drop iptables rule on node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Important note: Step [5] is necessary to reproduce _this_ bug. If we
miss step [5], the read command (step 6) will hang for a long time and
this commit doesn't help, because there will be not long connect() to
unreachable host, but long sendmsg() to unreachable host, which should
be fixed by enabling and adjusting keep-alive on the socket, which is a
thing for further patch set.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200903190301.367620-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-09-03 21:03:00 +02:00
|
|
|
reconnect_delay_timer_init(s, qemu_clock_get_ns(QEMU_CLOCK_REALTIME) +
|
|
|
|
s->reconnect_delay * NANOSECONDS_PER_SECOND);
|
|
|
|
}
|
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
nbd_reconnect_attempt(s);
|
|
|
|
|
|
|
|
while (nbd_client_connecting(s)) {
|
|
|
|
if (s->drained) {
|
|
|
|
bdrv_dec_in_flight(s->bs);
|
|
|
|
s->wait_drained_end = true;
|
|
|
|
while (s->drained) {
|
|
|
|
/*
|
|
|
|
* We may be entered once from nbd_client_attach_aio_context_bh
|
|
|
|
* and then from nbd_client_co_drain_end. So here is a loop.
|
|
|
|
*/
|
|
|
|
qemu_coroutine_yield();
|
|
|
|
}
|
|
|
|
bdrv_inc_in_flight(s->bs);
|
2020-07-27 20:47:50 +02:00
|
|
|
} else {
|
2021-05-17 12:05:47 +02:00
|
|
|
qemu_co_sleep_ns_wakeable(&s->reconnect_sleep,
|
|
|
|
QEMU_CLOCK_REALTIME, timeout);
|
2020-09-03 21:03:01 +02:00
|
|
|
if (s->drained) {
|
|
|
|
continue;
|
|
|
|
}
|
2020-07-27 20:47:50 +02:00
|
|
|
if (timeout < max_timeout) {
|
|
|
|
timeout *= 2;
|
|
|
|
}
|
2019-10-09 10:41:57 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
nbd_reconnect_attempt(s);
|
|
|
|
}
|
block/nbd: fix reconnect-delay
reconnect-delay has a design flaw: we handle it in the same loop where
we do connection attempt. So, reconnect-delay may be exceeded by
unpredictable time of connection attempt.
Let's instead use separate timer.
How to reproduce the bug:
1. Create an image on node1:
qemu-img create -f qcow2 xx 100M
2. Start NBD server on node1:
qemu-nbd xx
3. On node2 start qemu-io:
./build/qemu-io --image-opts \
driver=nbd,server.type=inet,server.host=192.168.100.5,server.port=10809,reconnect-delay=15
4. Type 'read 0 512' in qemu-io interface to check that connection
works
Be careful: you should make steps 5-7 in a short time, less than 15
seconds.
5. Kill nbd server on node1
6. Run 'read 0 512' in qemu-io interface again, to be sure that nbd
client goes to reconnect loop.
7. On node1 run the following command
sudo iptables -A INPUT -p tcp --dport 10809 -j DROP
This will make the connect() call of qemu-io at node2 take a long time.
And you'll see that read command in qemu-io will hang for a long time,
more than 15 seconds specified by reconnect-delay parameter. It's the
bug.
8. Don't forget to drop iptables rule on node1:
sudo iptables -D INPUT -p tcp --dport 10809 -j DROP
Important note: Step [5] is necessary to reproduce _this_ bug. If we
miss step [5], the read command (step 6) will hang for a long time and
this commit doesn't help, because there will be not long connect() to
unreachable host, but long sendmsg() to unreachable host, which should
be fixed by enabling and adjusting keep-alive on the socket, which is a
thing for further patch set.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20200903190301.367620-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2020-09-03 21:03:00 +02:00
|
|
|
|
|
|
|
reconnect_delay_timer_del(s);
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static coroutine_fn void nbd_connection_entry(void *opaque)
|
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
uint64_t i;
|
|
|
|
int ret = 0;
|
|
|
|
Error *local_err = NULL;
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
while (qatomic_load_acquire(&s->state) != NBD_CLIENT_QUIT) {
|
2019-06-11 12:27:19 +02:00
|
|
|
/*
|
|
|
|
* The NBD client can only really be considered idle when it has
|
|
|
|
* yielded from qio_channel_readv_all_eof(), waiting for data. This is
|
|
|
|
* the point where the additional scheduled coroutine entry happens
|
|
|
|
* after nbd_client_attach_aio_context().
|
|
|
|
*
|
|
|
|
* Therefore we keep an additional in_flight reference all the time and
|
|
|
|
* only drop it temporarily here.
|
|
|
|
*/
|
2019-10-09 10:41:57 +02:00
|
|
|
|
|
|
|
if (nbd_client_connecting(s)) {
|
|
|
|
nbd_co_reconnect_loop(s);
|
|
|
|
}
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) != NBD_CLIENT_CONNECTED) {
|
2019-10-09 10:41:57 +02:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:19 +02:00
|
|
|
assert(s->reply.handle == 0);
|
|
|
|
ret = nbd_receive_reply(s->bs, s->ioc, &s->reply, &local_err);
|
|
|
|
|
|
|
|
if (local_err) {
|
|
|
|
trace_nbd_read_reply_entry_fail(ret, error_get_pretty(local_err));
|
|
|
|
error_free(local_err);
|
2019-10-09 10:41:57 +02:00
|
|
|
local_err = NULL;
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
|
|
|
if (ret <= 0) {
|
2019-06-18 13:43:22 +02:00
|
|
|
nbd_channel_error(s, ret ? ret : -EIO);
|
2019-10-09 10:41:57 +02:00
|
|
|
continue;
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* There's no need for a mutex on the receive side, because the
|
|
|
|
* handler acts as a synchronization point and ensures that only
|
|
|
|
* one coroutine is called until the reply finishes.
|
|
|
|
*/
|
|
|
|
i = HANDLE_TO_INDEX(s, s->reply.handle);
|
|
|
|
if (i >= MAX_NBD_REQUESTS ||
|
|
|
|
!s->requests[i].coroutine ||
|
|
|
|
!s->requests[i].receiving ||
|
|
|
|
(nbd_reply_is_structured(&s->reply) && !s->info.structured_reply))
|
|
|
|
{
|
2019-06-18 13:43:22 +02:00
|
|
|
nbd_channel_error(s, -EINVAL);
|
2019-10-09 10:41:57 +02:00
|
|
|
continue;
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We're woken up again by the request itself. Note that there
|
|
|
|
* is no race between yielding and reentering connection_co. This
|
|
|
|
* is because:
|
|
|
|
*
|
|
|
|
* - if the request runs on the same AioContext, it is only
|
|
|
|
* entered after we yield
|
|
|
|
*
|
|
|
|
* - if the request runs on a different AioContext, reentering
|
|
|
|
* connection_co happens through a bottom half, which can only
|
|
|
|
* run after we yield.
|
|
|
|
*/
|
|
|
|
aio_co_wake(s->requests[i].coroutine);
|
|
|
|
qemu_coroutine_yield();
|
|
|
|
}
|
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
qemu_co_queue_restart_all(&s->free_sema);
|
2019-06-11 12:27:19 +02:00
|
|
|
nbd_recv_coroutines_wake_all(s);
|
|
|
|
bdrv_dec_in_flight(s->bs);
|
|
|
|
|
|
|
|
s->connection_co = NULL;
|
2019-10-09 10:41:57 +02:00
|
|
|
if (s->ioc) {
|
2020-09-03 21:02:59 +02:00
|
|
|
qio_channel_detach_aio_context(QIO_CHANNEL(s->ioc));
|
2020-12-28 16:08:44 +01:00
|
|
|
yank_unregister_function(BLOCKDEV_YANK_INSTANCE(s->bs->node_name),
|
|
|
|
nbd_yank, s->bs);
|
2019-10-09 10:41:57 +02:00
|
|
|
object_unref(OBJECT(s->sioc));
|
|
|
|
s->sioc = NULL;
|
|
|
|
object_unref(OBJECT(s->ioc));
|
|
|
|
s->ioc = NULL;
|
|
|
|
}
|
|
|
|
|
2020-01-22 17:45:28 +01:00
|
|
|
if (s->teardown_co) {
|
|
|
|
aio_co_wake(s->teardown_co);
|
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
aio_wait_kick();
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nbd_co_send_request(BlockDriverState *bs,
|
|
|
|
NBDRequest *request,
|
|
|
|
QEMUIOVector *qiov)
|
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-18 13:43:22 +02:00
|
|
|
int rc, i = -1;
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
qemu_co_mutex_lock(&s->send_mutex);
|
2019-10-09 10:41:57 +02:00
|
|
|
while (s->in_flight == MAX_NBD_REQUESTS || nbd_client_connecting_wait(s)) {
|
2019-06-11 12:27:19 +02:00
|
|
|
qemu_co_queue_wait(&s->free_sema, &s->send_mutex);
|
|
|
|
}
|
2019-06-18 13:43:22 +02:00
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) != NBD_CLIENT_CONNECTED) {
|
2019-06-18 13:43:22 +02:00
|
|
|
rc = -EIO;
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:19 +02:00
|
|
|
s->in_flight++;
|
|
|
|
|
|
|
|
for (i = 0; i < MAX_NBD_REQUESTS; i++) {
|
|
|
|
if (s->requests[i].coroutine == NULL) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
g_assert(qemu_in_coroutine());
|
|
|
|
assert(i < MAX_NBD_REQUESTS);
|
|
|
|
|
|
|
|
s->requests[i].coroutine = qemu_coroutine_self();
|
|
|
|
s->requests[i].offset = request->from;
|
|
|
|
s->requests[i].receiving = false;
|
|
|
|
|
|
|
|
request->handle = INDEX_TO_HANDLE(s, i);
|
|
|
|
|
|
|
|
assert(s->ioc);
|
|
|
|
|
|
|
|
if (qiov) {
|
|
|
|
qio_channel_set_cork(s->ioc, true);
|
|
|
|
rc = nbd_send_request(s->ioc, request);
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) == NBD_CLIENT_CONNECTED &&
|
|
|
|
rc >= 0) {
|
2019-06-11 12:27:19 +02:00
|
|
|
if (qio_channel_writev_all(s->ioc, qiov->iov, qiov->niov,
|
|
|
|
NULL) < 0) {
|
|
|
|
rc = -EIO;
|
|
|
|
}
|
|
|
|
} else if (rc >= 0) {
|
|
|
|
rc = -EIO;
|
|
|
|
}
|
|
|
|
qio_channel_set_cork(s->ioc, false);
|
|
|
|
} else {
|
|
|
|
rc = nbd_send_request(s->ioc, request);
|
|
|
|
}
|
|
|
|
|
|
|
|
err:
|
|
|
|
if (rc < 0) {
|
2019-06-18 13:43:22 +02:00
|
|
|
nbd_channel_error(s, rc);
|
|
|
|
if (i != -1) {
|
|
|
|
s->requests[i].coroutine = NULL;
|
|
|
|
s->in_flight--;
|
|
|
|
}
|
2019-10-09 10:41:57 +02:00
|
|
|
if (s->in_flight == 0 && s->wait_in_flight) {
|
|
|
|
aio_co_wake(s->connection_co);
|
|
|
|
} else {
|
|
|
|
qemu_co_queue_next(&s->free_sema);
|
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
|
|
|
qemu_co_mutex_unlock(&s->send_mutex);
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline uint16_t payload_advance16(uint8_t **payload)
|
|
|
|
{
|
|
|
|
*payload += 2;
|
|
|
|
return lduw_be_p(*payload - 2);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline uint32_t payload_advance32(uint8_t **payload)
|
|
|
|
{
|
|
|
|
*payload += 4;
|
|
|
|
return ldl_be_p(*payload - 4);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline uint64_t payload_advance64(uint8_t **payload)
|
|
|
|
{
|
|
|
|
*payload += 8;
|
|
|
|
return ldq_be_p(*payload - 8);
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
static int nbd_parse_offset_hole_payload(BDRVNBDState *s,
|
2019-06-11 12:27:19 +02:00
|
|
|
NBDStructuredReplyChunk *chunk,
|
|
|
|
uint8_t *payload, uint64_t orig_offset,
|
|
|
|
QEMUIOVector *qiov, Error **errp)
|
|
|
|
{
|
|
|
|
uint64_t offset;
|
|
|
|
uint32_t hole_size;
|
|
|
|
|
|
|
|
if (chunk->length != sizeof(offset) + sizeof(hole_size)) {
|
|
|
|
error_setg(errp, "Protocol error: invalid payload for "
|
|
|
|
"NBD_REPLY_TYPE_OFFSET_HOLE");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
offset = payload_advance64(&payload);
|
|
|
|
hole_size = payload_advance32(&payload);
|
|
|
|
|
|
|
|
if (!hole_size || offset < orig_offset || hole_size > qiov->size ||
|
|
|
|
offset > orig_offset + qiov->size - hole_size) {
|
|
|
|
error_setg(errp, "Protocol error: server sent chunk exceeding requested"
|
|
|
|
" region");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
2019-06-11 12:27:20 +02:00
|
|
|
if (s->info.min_block &&
|
|
|
|
!QEMU_IS_ALIGNED(hole_size, s->info.min_block)) {
|
2019-06-11 12:27:19 +02:00
|
|
|
trace_nbd_structured_read_compliance("hole");
|
|
|
|
}
|
|
|
|
|
|
|
|
qemu_iovec_memset(qiov, offset - orig_offset, 0, hole_size);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nbd_parse_blockstatus_payload
|
|
|
|
* Based on our request, we expect only one extent in reply, for the
|
|
|
|
* base:allocation context.
|
|
|
|
*/
|
2019-06-11 12:27:20 +02:00
|
|
|
static int nbd_parse_blockstatus_payload(BDRVNBDState *s,
|
2019-06-11 12:27:19 +02:00
|
|
|
NBDStructuredReplyChunk *chunk,
|
|
|
|
uint8_t *payload, uint64_t orig_length,
|
|
|
|
NBDExtent *extent, Error **errp)
|
|
|
|
{
|
|
|
|
uint32_t context_id;
|
|
|
|
|
|
|
|
/* The server succeeded, so it must have sent [at least] one extent */
|
|
|
|
if (chunk->length < sizeof(context_id) + sizeof(*extent)) {
|
|
|
|
error_setg(errp, "Protocol error: invalid payload for "
|
|
|
|
"NBD_REPLY_TYPE_BLOCK_STATUS");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
context_id = payload_advance32(&payload);
|
2019-06-11 12:27:20 +02:00
|
|
|
if (s->info.context_id != context_id) {
|
2019-06-11 12:27:19 +02:00
|
|
|
error_setg(errp, "Protocol error: unexpected context id %d for "
|
|
|
|
"NBD_REPLY_TYPE_BLOCK_STATUS, when negotiated context "
|
|
|
|
"id is %d", context_id,
|
2019-06-11 12:27:20 +02:00
|
|
|
s->info.context_id);
|
2019-06-11 12:27:19 +02:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
extent->length = payload_advance32(&payload);
|
|
|
|
extent->flags = payload_advance32(&payload);
|
|
|
|
|
|
|
|
if (extent->length == 0) {
|
|
|
|
error_setg(errp, "Protocol error: server sent status chunk with "
|
|
|
|
"zero length");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* A server sending unaligned block status is in violation of the
|
|
|
|
* protocol, but as qemu-nbd 3.1 is such a server (at least for
|
|
|
|
* POSIX files that are not a multiple of 512 bytes, since qemu
|
|
|
|
* rounds files up to 512-byte multiples but lseek(SEEK_HOLE)
|
|
|
|
* still sees an implicit hole beyond the real EOF), it's nicer to
|
|
|
|
* work around the misbehaving server. If the request included
|
|
|
|
* more than the final unaligned block, truncate it back to an
|
|
|
|
* aligned result; if the request was only the final block, round
|
|
|
|
* up to the full block and change the status to fully-allocated
|
|
|
|
* (always a safe status, even if it loses information).
|
|
|
|
*/
|
2019-06-11 12:27:20 +02:00
|
|
|
if (s->info.min_block && !QEMU_IS_ALIGNED(extent->length,
|
|
|
|
s->info.min_block)) {
|
2019-06-11 12:27:19 +02:00
|
|
|
trace_nbd_parse_blockstatus_compliance("extent length is unaligned");
|
2019-06-11 12:27:20 +02:00
|
|
|
if (extent->length > s->info.min_block) {
|
2019-06-11 12:27:19 +02:00
|
|
|
extent->length = QEMU_ALIGN_DOWN(extent->length,
|
2019-06-11 12:27:20 +02:00
|
|
|
s->info.min_block);
|
2019-06-11 12:27:19 +02:00
|
|
|
} else {
|
2019-06-11 12:27:20 +02:00
|
|
|
extent->length = s->info.min_block;
|
2019-06-11 12:27:19 +02:00
|
|
|
extent->flags = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We used NBD_CMD_FLAG_REQ_ONE, so the server should not have
|
|
|
|
* sent us any more than one extent, nor should it have included
|
|
|
|
* status beyond our request in that extent. However, it's easy
|
|
|
|
* enough to ignore the server's noncompliance without killing the
|
|
|
|
* connection; just ignore trailing extents, and clamp things to
|
|
|
|
* the length of our request.
|
|
|
|
*/
|
|
|
|
if (chunk->length > sizeof(context_id) + sizeof(*extent)) {
|
|
|
|
trace_nbd_parse_blockstatus_compliance("more than one extent");
|
|
|
|
}
|
|
|
|
if (extent->length > orig_length) {
|
|
|
|
extent->length = orig_length;
|
|
|
|
trace_nbd_parse_blockstatus_compliance("extent length too large");
|
|
|
|
}
|
|
|
|
|
2020-10-27 06:05:55 +01:00
|
|
|
/*
|
|
|
|
* HACK: if we are using x-dirty-bitmaps to access
|
|
|
|
* qemu:allocation-depth, treat all depths > 2 the same as 2,
|
|
|
|
* since nbd_client_co_block_status is only expecting the low two
|
|
|
|
* bits to be set.
|
|
|
|
*/
|
|
|
|
if (s->alloc_depth && extent->flags > 2) {
|
|
|
|
extent->flags = 2;
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:19 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nbd_parse_error_payload
|
|
|
|
* on success @errp contains message describing nbd error reply
|
|
|
|
*/
|
|
|
|
static int nbd_parse_error_payload(NBDStructuredReplyChunk *chunk,
|
|
|
|
uint8_t *payload, int *request_ret,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
uint32_t error;
|
|
|
|
uint16_t message_size;
|
|
|
|
|
|
|
|
assert(chunk->type & (1 << 15));
|
|
|
|
|
|
|
|
if (chunk->length < sizeof(error) + sizeof(message_size)) {
|
|
|
|
error_setg(errp,
|
|
|
|
"Protocol error: invalid payload for structured error");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
error = nbd_errno_to_system_errno(payload_advance32(&payload));
|
|
|
|
if (error == 0) {
|
|
|
|
error_setg(errp, "Protocol error: server sent structured error chunk "
|
|
|
|
"with error = 0");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
*request_ret = -error;
|
|
|
|
message_size = payload_advance16(&payload);
|
|
|
|
|
|
|
|
if (message_size > chunk->length - sizeof(error) - sizeof(message_size)) {
|
|
|
|
error_setg(errp, "Protocol error: server sent structured error chunk "
|
|
|
|
"with incorrect message size");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* TODO: Add a trace point to mention the server complaint */
|
|
|
|
|
|
|
|
/* TODO handle ERROR_OFFSET */
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
static int nbd_co_receive_offset_data_payload(BDRVNBDState *s,
|
2019-06-11 12:27:19 +02:00
|
|
|
uint64_t orig_offset,
|
|
|
|
QEMUIOVector *qiov, Error **errp)
|
|
|
|
{
|
|
|
|
QEMUIOVector sub_qiov;
|
|
|
|
uint64_t offset;
|
|
|
|
size_t data_size;
|
|
|
|
int ret;
|
|
|
|
NBDStructuredReplyChunk *chunk = &s->reply.structured;
|
|
|
|
|
|
|
|
assert(nbd_reply_is_structured(&s->reply));
|
|
|
|
|
|
|
|
/* The NBD spec requires at least one byte of payload */
|
|
|
|
if (chunk->length <= sizeof(offset)) {
|
|
|
|
error_setg(errp, "Protocol error: invalid payload for "
|
|
|
|
"NBD_REPLY_TYPE_OFFSET_DATA");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (nbd_read64(s->ioc, &offset, "OFFSET_DATA offset", errp) < 0) {
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
|
|
|
|
data_size = chunk->length - sizeof(offset);
|
|
|
|
assert(data_size);
|
|
|
|
if (offset < orig_offset || data_size > qiov->size ||
|
|
|
|
offset > orig_offset + qiov->size - data_size) {
|
|
|
|
error_setg(errp, "Protocol error: server sent chunk exceeding requested"
|
|
|
|
" region");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
if (s->info.min_block && !QEMU_IS_ALIGNED(data_size, s->info.min_block)) {
|
|
|
|
trace_nbd_structured_read_compliance("data");
|
|
|
|
}
|
|
|
|
|
|
|
|
qemu_iovec_init(&sub_qiov, qiov->niov);
|
|
|
|
qemu_iovec_concat(&sub_qiov, qiov, offset - orig_offset, data_size);
|
|
|
|
ret = qio_channel_readv_all(s->ioc, sub_qiov.iov, sub_qiov.niov, errp);
|
|
|
|
qemu_iovec_destroy(&sub_qiov);
|
|
|
|
|
|
|
|
return ret < 0 ? -EIO : 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
#define NBD_MAX_MALLOC_PAYLOAD 1000
|
|
|
|
static coroutine_fn int nbd_co_receive_structured_payload(
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s, void **payload, Error **errp)
|
2019-06-11 12:27:19 +02:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
uint32_t len;
|
|
|
|
|
|
|
|
assert(nbd_reply_is_structured(&s->reply));
|
|
|
|
|
|
|
|
len = s->reply.structured.length;
|
|
|
|
|
|
|
|
if (len == 0) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (payload == NULL) {
|
|
|
|
error_setg(errp, "Unexpected structured payload");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (len > NBD_MAX_MALLOC_PAYLOAD) {
|
|
|
|
error_setg(errp, "Payload too large");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
*payload = g_new(char, len);
|
|
|
|
ret = nbd_read(s->ioc, *payload, len, "structured payload", errp);
|
|
|
|
if (ret < 0) {
|
|
|
|
g_free(*payload);
|
|
|
|
*payload = NULL;
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nbd_co_do_receive_one_chunk
|
|
|
|
* for simple reply:
|
|
|
|
* set request_ret to received reply error
|
|
|
|
* if qiov is not NULL: read payload to @qiov
|
|
|
|
* for structured reply chunk:
|
|
|
|
* if error chunk: read payload, set @request_ret, do not set @payload
|
|
|
|
* else if offset_data chunk: read payload data to @qiov, do not set @payload
|
|
|
|
* else: read payload to @payload
|
|
|
|
*
|
|
|
|
* If function fails, @errp contains corresponding error message, and the
|
|
|
|
* connection with the server is suspect. If it returns 0, then the
|
|
|
|
* transaction succeeded (although @request_ret may be a negative errno
|
|
|
|
* corresponding to the server's error reply), and errp is unchanged.
|
|
|
|
*/
|
|
|
|
static coroutine_fn int nbd_co_do_receive_one_chunk(
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s, uint64_t handle, bool only_structured,
|
2019-06-11 12:27:19 +02:00
|
|
|
int *request_ret, QEMUIOVector *qiov, void **payload, Error **errp)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
int i = HANDLE_TO_INDEX(s, handle);
|
|
|
|
void *local_payload = NULL;
|
|
|
|
NBDStructuredReplyChunk *chunk;
|
|
|
|
|
|
|
|
if (payload) {
|
|
|
|
*payload = NULL;
|
|
|
|
}
|
|
|
|
*request_ret = 0;
|
|
|
|
|
|
|
|
/* Wait until we're woken up by nbd_connection_entry. */
|
|
|
|
s->requests[i].receiving = true;
|
|
|
|
qemu_coroutine_yield();
|
|
|
|
s->requests[i].receiving = false;
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) != NBD_CLIENT_CONNECTED) {
|
2019-06-11 12:27:19 +02:00
|
|
|
error_setg(errp, "Connection closed");
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
assert(s->ioc);
|
|
|
|
|
|
|
|
assert(s->reply.handle == handle);
|
|
|
|
|
|
|
|
if (nbd_reply_is_simple(&s->reply)) {
|
|
|
|
if (only_structured) {
|
|
|
|
error_setg(errp, "Protocol error: simple reply when structured "
|
|
|
|
"reply chunk was expected");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
*request_ret = -nbd_errno_to_system_errno(s->reply.simple.error);
|
|
|
|
if (*request_ret < 0 || !qiov) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
return qio_channel_readv_all(s->ioc, qiov->iov, qiov->niov,
|
|
|
|
errp) < 0 ? -EIO : 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* handle structured reply chunk */
|
|
|
|
assert(s->info.structured_reply);
|
|
|
|
chunk = &s->reply.structured;
|
|
|
|
|
|
|
|
if (chunk->type == NBD_REPLY_TYPE_NONE) {
|
|
|
|
if (!(chunk->flags & NBD_REPLY_FLAG_DONE)) {
|
|
|
|
error_setg(errp, "Protocol error: NBD_REPLY_TYPE_NONE chunk without"
|
|
|
|
" NBD_REPLY_FLAG_DONE flag set");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
if (chunk->length) {
|
|
|
|
error_setg(errp, "Protocol error: NBD_REPLY_TYPE_NONE chunk with"
|
|
|
|
" nonzero length");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (chunk->type == NBD_REPLY_TYPE_OFFSET_DATA) {
|
|
|
|
if (!qiov) {
|
|
|
|
error_setg(errp, "Unexpected NBD_REPLY_TYPE_OFFSET_DATA chunk");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
return nbd_co_receive_offset_data_payload(s, s->requests[i].offset,
|
|
|
|
qiov, errp);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (nbd_reply_type_is_error(chunk->type)) {
|
|
|
|
payload = &local_payload;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = nbd_co_receive_structured_payload(s, payload, errp);
|
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (nbd_reply_type_is_error(chunk->type)) {
|
|
|
|
ret = nbd_parse_error_payload(chunk, local_payload, request_ret, errp);
|
|
|
|
g_free(local_payload);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nbd_co_receive_one_chunk
|
|
|
|
* Read reply, wake up connection_co and set s->quit if needed.
|
|
|
|
* Return value is a fatal error code or normal nbd reply error code
|
|
|
|
*/
|
|
|
|
static coroutine_fn int nbd_co_receive_one_chunk(
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s, uint64_t handle, bool only_structured,
|
2019-06-11 12:27:19 +02:00
|
|
|
int *request_ret, QEMUIOVector *qiov, NBDReply *reply, void **payload,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
int ret = nbd_co_do_receive_one_chunk(s, handle, only_structured,
|
|
|
|
request_ret, qiov, payload, errp);
|
|
|
|
|
|
|
|
if (ret < 0) {
|
2019-07-19 19:20:01 +02:00
|
|
|
memset(reply, 0, sizeof(*reply));
|
2019-06-18 13:43:22 +02:00
|
|
|
nbd_channel_error(s, ret);
|
2019-06-11 12:27:19 +02:00
|
|
|
} else {
|
|
|
|
/* For assert at loop start in nbd_connection_entry */
|
2019-07-19 19:20:01 +02:00
|
|
|
*reply = s->reply;
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
2019-10-09 10:41:57 +02:00
|
|
|
s->reply.handle = 0;
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
if (s->connection_co && !s->wait_in_flight) {
|
|
|
|
/*
|
|
|
|
* We must check s->wait_in_flight, because we may entered by
|
|
|
|
* nbd_recv_coroutines_wake_all(), in this case we should not
|
|
|
|
* wake connection_co here, it will woken by last request.
|
|
|
|
*/
|
2019-06-11 12:27:19 +02:00
|
|
|
aio_co_wake(s->connection_co);
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
typedef struct NBDReplyChunkIter {
|
|
|
|
int ret;
|
|
|
|
int request_ret;
|
|
|
|
Error *err;
|
|
|
|
bool done, only_structured;
|
|
|
|
} NBDReplyChunkIter;
|
|
|
|
|
|
|
|
static void nbd_iter_channel_error(NBDReplyChunkIter *iter,
|
|
|
|
int ret, Error **local_err)
|
|
|
|
{
|
2019-12-05 18:46:35 +01:00
|
|
|
assert(local_err && *local_err);
|
2019-06-11 12:27:19 +02:00
|
|
|
assert(ret < 0);
|
|
|
|
|
|
|
|
if (!iter->ret) {
|
|
|
|
iter->ret = ret;
|
|
|
|
error_propagate(&iter->err, *local_err);
|
|
|
|
} else {
|
|
|
|
error_free(*local_err);
|
|
|
|
}
|
|
|
|
|
|
|
|
*local_err = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void nbd_iter_request_error(NBDReplyChunkIter *iter, int ret)
|
|
|
|
{
|
|
|
|
assert(ret < 0);
|
|
|
|
|
|
|
|
if (!iter->request_ret) {
|
|
|
|
iter->request_ret = ret;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* NBD_FOREACH_REPLY_CHUNK
|
|
|
|
* The pointer stored in @payload requires g_free() to free it.
|
|
|
|
*/
|
|
|
|
#define NBD_FOREACH_REPLY_CHUNK(s, iter, handle, structured, \
|
|
|
|
qiov, reply, payload) \
|
|
|
|
for (iter = (NBDReplyChunkIter) { .only_structured = structured }; \
|
|
|
|
nbd_reply_chunk_iter_receive(s, &iter, handle, qiov, reply, payload);)
|
|
|
|
|
|
|
|
/*
|
|
|
|
* nbd_reply_chunk_iter_receive
|
|
|
|
* The pointer stored in @payload requires g_free() to free it.
|
|
|
|
*/
|
2019-06-11 12:27:20 +02:00
|
|
|
static bool nbd_reply_chunk_iter_receive(BDRVNBDState *s,
|
2019-06-11 12:27:19 +02:00
|
|
|
NBDReplyChunkIter *iter,
|
|
|
|
uint64_t handle,
|
|
|
|
QEMUIOVector *qiov, NBDReply *reply,
|
|
|
|
void **payload)
|
|
|
|
{
|
|
|
|
int ret, request_ret;
|
|
|
|
NBDReply local_reply;
|
|
|
|
NBDStructuredReplyChunk *chunk;
|
|
|
|
Error *local_err = NULL;
|
2020-12-28 16:08:44 +01:00
|
|
|
if (qatomic_load_acquire(&s->state) != NBD_CLIENT_CONNECTED) {
|
2019-06-11 12:27:19 +02:00
|
|
|
error_setg(&local_err, "Connection closed");
|
|
|
|
nbd_iter_channel_error(iter, -EIO, &local_err);
|
|
|
|
goto break_loop;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (iter->done) {
|
|
|
|
/* Previous iteration was last. */
|
|
|
|
goto break_loop;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (reply == NULL) {
|
|
|
|
reply = &local_reply;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = nbd_co_receive_one_chunk(s, handle, iter->only_structured,
|
|
|
|
&request_ret, qiov, reply, payload,
|
|
|
|
&local_err);
|
|
|
|
if (ret < 0) {
|
|
|
|
nbd_iter_channel_error(iter, ret, &local_err);
|
|
|
|
} else if (request_ret < 0) {
|
|
|
|
nbd_iter_request_error(iter, request_ret);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Do not execute the body of NBD_FOREACH_REPLY_CHUNK for simple reply. */
|
2020-12-28 16:08:44 +01:00
|
|
|
if (nbd_reply_is_simple(reply) ||
|
|
|
|
qatomic_load_acquire(&s->state) != NBD_CLIENT_CONNECTED) {
|
2019-06-11 12:27:19 +02:00
|
|
|
goto break_loop;
|
|
|
|
}
|
|
|
|
|
|
|
|
chunk = &reply->structured;
|
|
|
|
iter->only_structured = true;
|
|
|
|
|
|
|
|
if (chunk->type == NBD_REPLY_TYPE_NONE) {
|
|
|
|
/* NBD_REPLY_FLAG_DONE is already checked in nbd_co_receive_one_chunk */
|
|
|
|
assert(chunk->flags & NBD_REPLY_FLAG_DONE);
|
|
|
|
goto break_loop;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (chunk->flags & NBD_REPLY_FLAG_DONE) {
|
|
|
|
/* This iteration is last. */
|
|
|
|
iter->done = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Execute the loop body */
|
|
|
|
return true;
|
|
|
|
|
|
|
|
break_loop:
|
|
|
|
s->requests[HANDLE_TO_INDEX(s, handle)].coroutine = NULL;
|
|
|
|
|
|
|
|
qemu_co_mutex_lock(&s->send_mutex);
|
|
|
|
s->in_flight--;
|
2019-10-09 10:41:57 +02:00
|
|
|
if (s->in_flight == 0 && s->wait_in_flight) {
|
|
|
|
aio_co_wake(s->connection_co);
|
|
|
|
} else {
|
|
|
|
qemu_co_queue_next(&s->free_sema);
|
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
qemu_co_mutex_unlock(&s->send_mutex);
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
static int nbd_co_receive_return_code(BDRVNBDState *s, uint64_t handle,
|
2019-06-11 12:27:19 +02:00
|
|
|
int *request_ret, Error **errp)
|
|
|
|
{
|
|
|
|
NBDReplyChunkIter iter;
|
|
|
|
|
|
|
|
NBD_FOREACH_REPLY_CHUNK(s, iter, handle, false, NULL, NULL, NULL) {
|
|
|
|
/* nbd_reply_chunk_iter_receive does all the work */
|
|
|
|
}
|
|
|
|
|
|
|
|
error_propagate(errp, iter.err);
|
|
|
|
*request_ret = iter.request_ret;
|
|
|
|
return iter.ret;
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
static int nbd_co_receive_cmdread_reply(BDRVNBDState *s, uint64_t handle,
|
2019-06-11 12:27:19 +02:00
|
|
|
uint64_t offset, QEMUIOVector *qiov,
|
|
|
|
int *request_ret, Error **errp)
|
|
|
|
{
|
|
|
|
NBDReplyChunkIter iter;
|
|
|
|
NBDReply reply;
|
|
|
|
void *payload = NULL;
|
|
|
|
Error *local_err = NULL;
|
|
|
|
|
|
|
|
NBD_FOREACH_REPLY_CHUNK(s, iter, handle, s->info.structured_reply,
|
|
|
|
qiov, &reply, &payload)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
NBDStructuredReplyChunk *chunk = &reply.structured;
|
|
|
|
|
|
|
|
assert(nbd_reply_is_structured(&reply));
|
|
|
|
|
|
|
|
switch (chunk->type) {
|
|
|
|
case NBD_REPLY_TYPE_OFFSET_DATA:
|
|
|
|
/*
|
|
|
|
* special cased in nbd_co_receive_one_chunk, data is already
|
|
|
|
* in qiov
|
|
|
|
*/
|
|
|
|
break;
|
|
|
|
case NBD_REPLY_TYPE_OFFSET_HOLE:
|
|
|
|
ret = nbd_parse_offset_hole_payload(s, &reply.structured, payload,
|
|
|
|
offset, qiov, &local_err);
|
|
|
|
if (ret < 0) {
|
2019-06-18 13:43:22 +02:00
|
|
|
nbd_channel_error(s, ret);
|
2019-06-11 12:27:19 +02:00
|
|
|
nbd_iter_channel_error(&iter, ret, &local_err);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
if (!nbd_reply_type_is_error(chunk->type)) {
|
|
|
|
/* not allowed reply type */
|
2019-06-18 13:43:22 +02:00
|
|
|
nbd_channel_error(s, -EINVAL);
|
2019-06-11 12:27:19 +02:00
|
|
|
error_setg(&local_err,
|
|
|
|
"Unexpected reply type: %d (%s) for CMD_READ",
|
|
|
|
chunk->type, nbd_reply_type_lookup(chunk->type));
|
|
|
|
nbd_iter_channel_error(&iter, -EINVAL, &local_err);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
g_free(payload);
|
|
|
|
payload = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
error_propagate(errp, iter.err);
|
|
|
|
*request_ret = iter.request_ret;
|
|
|
|
return iter.ret;
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
static int nbd_co_receive_blockstatus_reply(BDRVNBDState *s,
|
2019-06-11 12:27:19 +02:00
|
|
|
uint64_t handle, uint64_t length,
|
|
|
|
NBDExtent *extent,
|
|
|
|
int *request_ret, Error **errp)
|
|
|
|
{
|
|
|
|
NBDReplyChunkIter iter;
|
|
|
|
NBDReply reply;
|
|
|
|
void *payload = NULL;
|
|
|
|
Error *local_err = NULL;
|
|
|
|
bool received = false;
|
|
|
|
|
|
|
|
assert(!extent->length);
|
|
|
|
NBD_FOREACH_REPLY_CHUNK(s, iter, handle, false, NULL, &reply, &payload) {
|
|
|
|
int ret;
|
|
|
|
NBDStructuredReplyChunk *chunk = &reply.structured;
|
|
|
|
|
|
|
|
assert(nbd_reply_is_structured(&reply));
|
|
|
|
|
|
|
|
switch (chunk->type) {
|
|
|
|
case NBD_REPLY_TYPE_BLOCK_STATUS:
|
|
|
|
if (received) {
|
2019-06-18 13:43:22 +02:00
|
|
|
nbd_channel_error(s, -EINVAL);
|
2019-06-11 12:27:19 +02:00
|
|
|
error_setg(&local_err, "Several BLOCK_STATUS chunks in reply");
|
|
|
|
nbd_iter_channel_error(&iter, -EINVAL, &local_err);
|
|
|
|
}
|
|
|
|
received = true;
|
|
|
|
|
|
|
|
ret = nbd_parse_blockstatus_payload(s, &reply.structured,
|
|
|
|
payload, length, extent,
|
|
|
|
&local_err);
|
|
|
|
if (ret < 0) {
|
2019-06-18 13:43:22 +02:00
|
|
|
nbd_channel_error(s, ret);
|
2019-06-11 12:27:19 +02:00
|
|
|
nbd_iter_channel_error(&iter, ret, &local_err);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
if (!nbd_reply_type_is_error(chunk->type)) {
|
2019-06-18 13:43:22 +02:00
|
|
|
nbd_channel_error(s, -EINVAL);
|
2019-06-11 12:27:19 +02:00
|
|
|
error_setg(&local_err,
|
|
|
|
"Unexpected reply type: %d (%s) "
|
|
|
|
"for CMD_BLOCK_STATUS",
|
|
|
|
chunk->type, nbd_reply_type_lookup(chunk->type));
|
|
|
|
nbd_iter_channel_error(&iter, -EINVAL, &local_err);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
g_free(payload);
|
|
|
|
payload = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!extent->length && !iter.request_ret) {
|
|
|
|
error_setg(&local_err, "Server did not reply with any status extents");
|
|
|
|
nbd_iter_channel_error(&iter, -EIO, &local_err);
|
|
|
|
}
|
|
|
|
|
|
|
|
error_propagate(errp, iter.err);
|
|
|
|
*request_ret = iter.request_ret;
|
|
|
|
return iter.ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nbd_co_request(BlockDriverState *bs, NBDRequest *request,
|
|
|
|
QEMUIOVector *write_qiov)
|
|
|
|
{
|
|
|
|
int ret, request_ret;
|
|
|
|
Error *local_err = NULL;
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
assert(request->type != NBD_CMD_READ);
|
|
|
|
if (write_qiov) {
|
|
|
|
assert(request->type == NBD_CMD_WRITE);
|
|
|
|
assert(request->len == iov_size(write_qiov->iov, write_qiov->niov));
|
|
|
|
} else {
|
|
|
|
assert(request->type != NBD_CMD_WRITE);
|
|
|
|
}
|
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
do {
|
|
|
|
ret = nbd_co_send_request(bs, request, write_qiov);
|
|
|
|
if (ret < 0) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = nbd_co_receive_return_code(s, request->handle,
|
|
|
|
&request_ret, &local_err);
|
|
|
|
if (local_err) {
|
|
|
|
trace_nbd_co_request_fail(request->from, request->len,
|
|
|
|
request->handle, request->flags,
|
|
|
|
request->type,
|
|
|
|
nbd_cmd_lookup(request->type),
|
|
|
|
ret, error_get_pretty(local_err));
|
|
|
|
error_free(local_err);
|
|
|
|
local_err = NULL;
|
|
|
|
}
|
|
|
|
} while (ret < 0 && nbd_client_connecting_wait(s));
|
|
|
|
|
2019-06-11 12:27:19 +02:00
|
|
|
return ret ? ret : request_ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nbd_client_co_preadv(BlockDriverState *bs, uint64_t offset,
|
|
|
|
uint64_t bytes, QEMUIOVector *qiov, int flags)
|
|
|
|
{
|
|
|
|
int ret, request_ret;
|
|
|
|
Error *local_err = NULL;
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
NBDRequest request = {
|
|
|
|
.type = NBD_CMD_READ,
|
|
|
|
.from = offset,
|
|
|
|
.len = bytes,
|
|
|
|
};
|
|
|
|
|
|
|
|
assert(bytes <= NBD_MAX_BUFFER_SIZE);
|
|
|
|
assert(!flags);
|
|
|
|
|
|
|
|
if (!bytes) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Work around the fact that the block layer doesn't do
|
|
|
|
* byte-accurate sizing yet - if the read exceeds the server's
|
|
|
|
* advertised size because the block layer rounded size up, then
|
|
|
|
* truncate the request to the server and tail-pad with zero.
|
|
|
|
*/
|
2019-06-11 12:27:20 +02:00
|
|
|
if (offset >= s->info.size) {
|
2019-06-11 12:27:19 +02:00
|
|
|
assert(bytes < BDRV_SECTOR_SIZE);
|
|
|
|
qemu_iovec_memset(qiov, 0, 0, bytes);
|
|
|
|
return 0;
|
|
|
|
}
|
2019-06-11 12:27:20 +02:00
|
|
|
if (offset + bytes > s->info.size) {
|
|
|
|
uint64_t slop = offset + bytes - s->info.size;
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
assert(slop < BDRV_SECTOR_SIZE);
|
|
|
|
qemu_iovec_memset(qiov, bytes - slop, 0, slop);
|
|
|
|
request.len -= slop;
|
|
|
|
}
|
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
do {
|
|
|
|
ret = nbd_co_send_request(bs, &request, NULL);
|
|
|
|
if (ret < 0) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = nbd_co_receive_cmdread_reply(s, request.handle, offset, qiov,
|
|
|
|
&request_ret, &local_err);
|
|
|
|
if (local_err) {
|
|
|
|
trace_nbd_co_request_fail(request.from, request.len, request.handle,
|
|
|
|
request.flags, request.type,
|
|
|
|
nbd_cmd_lookup(request.type),
|
|
|
|
ret, error_get_pretty(local_err));
|
|
|
|
error_free(local_err);
|
|
|
|
local_err = NULL;
|
|
|
|
}
|
|
|
|
} while (ret < 0 && nbd_client_connecting_wait(s));
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
return ret ? ret : request_ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nbd_client_co_pwritev(BlockDriverState *bs, uint64_t offset,
|
|
|
|
uint64_t bytes, QEMUIOVector *qiov, int flags)
|
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
NBDRequest request = {
|
|
|
|
.type = NBD_CMD_WRITE,
|
|
|
|
.from = offset,
|
|
|
|
.len = bytes,
|
|
|
|
};
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
assert(!(s->info.flags & NBD_FLAG_READ_ONLY));
|
2019-06-11 12:27:19 +02:00
|
|
|
if (flags & BDRV_REQ_FUA) {
|
2019-06-11 12:27:20 +02:00
|
|
|
assert(s->info.flags & NBD_FLAG_SEND_FUA);
|
2019-06-11 12:27:19 +02:00
|
|
|
request.flags |= NBD_CMD_FLAG_FUA;
|
|
|
|
}
|
|
|
|
|
|
|
|
assert(bytes <= NBD_MAX_BUFFER_SIZE);
|
|
|
|
|
|
|
|
if (!bytes) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
return nbd_co_request(bs, &request, qiov);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nbd_client_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,
|
|
|
|
int bytes, BdrvRequestFlags flags)
|
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
NBDRequest request = {
|
|
|
|
.type = NBD_CMD_WRITE_ZEROES,
|
|
|
|
.from = offset,
|
|
|
|
.len = bytes,
|
|
|
|
};
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
assert(!(s->info.flags & NBD_FLAG_READ_ONLY));
|
|
|
|
if (!(s->info.flags & NBD_FLAG_SEND_WRITE_ZEROES)) {
|
2019-06-11 12:27:19 +02:00
|
|
|
return -ENOTSUP;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (flags & BDRV_REQ_FUA) {
|
2019-06-11 12:27:20 +02:00
|
|
|
assert(s->info.flags & NBD_FLAG_SEND_FUA);
|
2019-06-11 12:27:19 +02:00
|
|
|
request.flags |= NBD_CMD_FLAG_FUA;
|
|
|
|
}
|
|
|
|
if (!(flags & BDRV_REQ_MAY_UNMAP)) {
|
|
|
|
request.flags |= NBD_CMD_FLAG_NO_HOLE;
|
|
|
|
}
|
2019-08-23 16:37:24 +02:00
|
|
|
if (flags & BDRV_REQ_NO_FALLBACK) {
|
|
|
|
assert(s->info.flags & NBD_FLAG_SEND_FAST_ZERO);
|
|
|
|
request.flags |= NBD_CMD_FLAG_FAST_ZERO;
|
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
if (!bytes) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
return nbd_co_request(bs, &request, NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nbd_client_co_flush(BlockDriverState *bs)
|
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
NBDRequest request = { .type = NBD_CMD_FLUSH };
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
if (!(s->info.flags & NBD_FLAG_SEND_FLUSH)) {
|
2019-06-11 12:27:19 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
request.from = 0;
|
|
|
|
request.len = 0;
|
|
|
|
|
|
|
|
return nbd_co_request(bs, &request, NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int nbd_client_co_pdiscard(BlockDriverState *bs, int64_t offset,
|
|
|
|
int bytes)
|
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
NBDRequest request = {
|
|
|
|
.type = NBD_CMD_TRIM,
|
|
|
|
.from = offset,
|
|
|
|
.len = bytes,
|
|
|
|
};
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
assert(!(s->info.flags & NBD_FLAG_READ_ONLY));
|
|
|
|
if (!(s->info.flags & NBD_FLAG_SEND_TRIM) || !bytes) {
|
2019-06-11 12:27:19 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
return nbd_co_request(bs, &request, NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int coroutine_fn nbd_client_co_block_status(
|
|
|
|
BlockDriverState *bs, bool want_zero, int64_t offset, int64_t bytes,
|
|
|
|
int64_t *pnum, int64_t *map, BlockDriverState **file)
|
|
|
|
{
|
|
|
|
int ret, request_ret;
|
|
|
|
NBDExtent extent = { 0 };
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
Error *local_err = NULL;
|
|
|
|
|
|
|
|
NBDRequest request = {
|
|
|
|
.type = NBD_CMD_BLOCK_STATUS,
|
|
|
|
.from = offset,
|
2020-04-01 17:01:07 +02:00
|
|
|
.len = MIN(QEMU_ALIGN_DOWN(INT_MAX, bs->bl.request_alignment),
|
2019-06-11 12:27:20 +02:00
|
|
|
MIN(bytes, s->info.size - offset)),
|
2019-06-11 12:27:19 +02:00
|
|
|
.flags = NBD_CMD_FLAG_REQ_ONE,
|
|
|
|
};
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
if (!s->info.base_allocation) {
|
2019-06-11 12:27:19 +02:00
|
|
|
*pnum = bytes;
|
|
|
|
*map = offset;
|
|
|
|
*file = bs;
|
|
|
|
return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Work around the fact that the block layer doesn't do
|
|
|
|
* byte-accurate sizing yet - if the status request exceeds the
|
|
|
|
* server's advertised size because the block layer rounded size
|
|
|
|
* up, we truncated the request to the server (above), or are
|
|
|
|
* called on just the hole.
|
|
|
|
*/
|
2019-06-11 12:27:20 +02:00
|
|
|
if (offset >= s->info.size) {
|
2019-06-11 12:27:19 +02:00
|
|
|
*pnum = bytes;
|
|
|
|
assert(bytes < BDRV_SECTOR_SIZE);
|
|
|
|
/* Intentionally don't report offset_valid for the hole */
|
|
|
|
return BDRV_BLOCK_ZERO;
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
if (s->info.min_block) {
|
|
|
|
assert(QEMU_IS_ALIGNED(request.len, s->info.min_block));
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
2019-10-09 10:41:57 +02:00
|
|
|
do {
|
|
|
|
ret = nbd_co_send_request(bs, &request, NULL);
|
|
|
|
if (ret < 0) {
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = nbd_co_receive_blockstatus_reply(s, request.handle, bytes,
|
|
|
|
&extent, &request_ret,
|
|
|
|
&local_err);
|
|
|
|
if (local_err) {
|
|
|
|
trace_nbd_co_request_fail(request.from, request.len, request.handle,
|
|
|
|
request.flags, request.type,
|
|
|
|
nbd_cmd_lookup(request.type),
|
|
|
|
ret, error_get_pretty(local_err));
|
|
|
|
error_free(local_err);
|
|
|
|
local_err = NULL;
|
|
|
|
}
|
|
|
|
} while (ret < 0 && nbd_client_connecting_wait(s));
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
if (ret < 0 || request_ret < 0) {
|
|
|
|
return ret ? ret : request_ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
assert(extent.length);
|
|
|
|
*pnum = extent.length;
|
|
|
|
*map = offset;
|
|
|
|
*file = bs;
|
|
|
|
return (extent.flags & NBD_STATE_HOLE ? 0 : BDRV_BLOCK_DATA) |
|
|
|
|
(extent.flags & NBD_STATE_ZERO ? BDRV_BLOCK_ZERO : 0) |
|
|
|
|
BDRV_BLOCK_OFFSET_VALID;
|
|
|
|
}
|
|
|
|
|
2019-09-30 23:38:20 +02:00
|
|
|
static int nbd_client_reopen_prepare(BDRVReopenState *state,
|
|
|
|
BlockReopenQueue *queue, Error **errp)
|
|
|
|
{
|
|
|
|
BDRVNBDState *s = (BDRVNBDState *)state->bs->opaque;
|
|
|
|
|
|
|
|
if ((state->flags & BDRV_O_RDWR) && (s->info.flags & NBD_FLAG_READ_ONLY)) {
|
|
|
|
error_setg(errp, "Can't reopen read-only NBD mount as read/write");
|
|
|
|
return -EACCES;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
static void nbd_yank(void *opaque)
|
|
|
|
{
|
|
|
|
BlockDriverState *bs = opaque;
|
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
|
|
|
|
|
|
|
qatomic_store_release(&s->state, NBD_CLIENT_QUIT);
|
|
|
|
qio_channel_shutdown(QIO_CHANNEL(s->sioc), QIO_CHANNEL_SHUTDOWN_BOTH, NULL);
|
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:19 +02:00
|
|
|
static void nbd_client_close(BlockDriverState *bs)
|
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
NBDRequest request = { .type = NBD_CMD_DISC };
|
|
|
|
|
2019-10-09 10:41:57 +02:00
|
|
|
if (s->ioc) {
|
|
|
|
nbd_send_request(s->ioc, &request);
|
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
nbd_teardown_connection(bs);
|
|
|
|
}
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
static int nbd_establish_connection(BlockDriverState *bs,
|
|
|
|
SocketAddress *saddr,
|
|
|
|
Error **errp)
|
2019-06-11 12:27:19 +02:00
|
|
|
{
|
nbd: Use ERRP_GUARD()
If we want to check error after errp-function call, we need to
introduce local_err and then propagate it to errp. Instead, use
the ERRP_GUARD() macro, benefits are:
1. No need of explicit error_propagate call
2. No need of explicit local_err variable: use errp directly
3. ERRP_GUARD() leaves errp as is if it's not NULL or
&error_fatal, this means that we don't break error_abort
(we'll abort on error_set, not on error_propagate)
If we want to add some info to errp (by error_prepend() or
error_append_hint()), we must use the ERRP_GUARD() macro.
Otherwise, this info will not be added when errp == &error_fatal
(the program will exit prior to the error_append_hint() or
error_prepend() call). Fix several such cases, e.g. in nbd_read().
This commit is generated by command
sed -n '/^Network Block Device (NBD)$/,/^$/{s/^F: //p}' \
MAINTAINERS | \
xargs git ls-files | grep '\.[hc]$' | \
xargs spatch \
--sp-file scripts/coccinelle/errp-guard.cocci \
--macro-file scripts/cocci-macro-file.h \
--in-place --no-show-diff --max-width 80
Reported-by: Kevin Wolf <kwolf@redhat.com>
Reported-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[Commit message tweaked]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20200707165037.1026246-8-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[ERRP_AUTO_PROPAGATE() renamed to ERRP_GUARD(), and
auto-propagated-errp.cocci to errp-guard.cocci. Commit message
tweaked again.]
2020-07-07 18:50:36 +02:00
|
|
|
ERRP_GUARD();
|
2020-12-28 16:08:44 +01:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
s->sioc = qio_channel_socket_new();
|
|
|
|
qio_channel_set_name(QIO_CHANNEL(s->sioc), "nbd-client");
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
qio_channel_socket_connect_sync(s->sioc, saddr, errp);
|
nbd: Use ERRP_GUARD()
If we want to check error after errp-function call, we need to
introduce local_err and then propagate it to errp. Instead, use
the ERRP_GUARD() macro, benefits are:
1. No need of explicit error_propagate call
2. No need of explicit local_err variable: use errp directly
3. ERRP_GUARD() leaves errp as is if it's not NULL or
&error_fatal, this means that we don't break error_abort
(we'll abort on error_set, not on error_propagate)
If we want to add some info to errp (by error_prepend() or
error_append_hint()), we must use the ERRP_GUARD() macro.
Otherwise, this info will not be added when errp == &error_fatal
(the program will exit prior to the error_append_hint() or
error_prepend() call). Fix several such cases, e.g. in nbd_read().
This commit is generated by command
sed -n '/^Network Block Device (NBD)$/,/^$/{s/^F: //p}' \
MAINTAINERS | \
xargs git ls-files | grep '\.[hc]$' | \
xargs spatch \
--sp-file scripts/coccinelle/errp-guard.cocci \
--macro-file scripts/cocci-macro-file.h \
--in-place --no-show-diff --max-width 80
Reported-by: Kevin Wolf <kwolf@redhat.com>
Reported-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[Commit message tweaked]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20200707165037.1026246-8-armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[ERRP_AUTO_PROPAGATE() renamed to ERRP_GUARD(), and
auto-propagated-errp.cocci to errp-guard.cocci. Commit message
tweaked again.]
2020-07-07 18:50:36 +02:00
|
|
|
if (*errp) {
|
2020-12-28 16:08:44 +01:00
|
|
|
object_unref(OBJECT(s->sioc));
|
|
|
|
s->sioc = NULL;
|
|
|
|
return -1;
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
yank_register_function(BLOCKDEV_YANK_INSTANCE(bs->node_name), nbd_yank, bs);
|
|
|
|
qio_channel_set_delay(QIO_CHANNEL(s->sioc), false);
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
return 0;
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
/* nbd_client_handshake takes ownership on s->sioc. On failure it's unref'ed. */
|
|
|
|
static int nbd_client_handshake(BlockDriverState *bs, Error **errp)
|
2019-06-11 12:27:19 +02:00
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
2019-06-18 13:43:21 +02:00
|
|
|
AioContext *aio_context = bdrv_get_aio_context(bs);
|
2019-06-11 12:27:19 +02:00
|
|
|
int ret;
|
|
|
|
|
2020-07-27 20:47:47 +02:00
|
|
|
trace_nbd_client_handshake(s->export);
|
2020-12-28 16:08:44 +01:00
|
|
|
qio_channel_set_blocking(QIO_CHANNEL(s->sioc), false, NULL);
|
|
|
|
qio_channel_attach_aio_context(QIO_CHANNEL(s->sioc), aio_context);
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
s->info.request_sizes = true;
|
|
|
|
s->info.structured_reply = true;
|
|
|
|
s->info.base_allocation = true;
|
2019-06-18 13:43:24 +02:00
|
|
|
s->info.x_dirty_bitmap = g_strdup(s->x_dirty_bitmap);
|
|
|
|
s->info.name = g_strdup(s->export ?: "");
|
2020-12-28 16:08:44 +01:00
|
|
|
ret = nbd_receive_negotiate(aio_context, QIO_CHANNEL(s->sioc), s->tlscreds,
|
2019-06-18 13:43:24 +02:00
|
|
|
s->hostname, &s->ioc, &s->info, errp);
|
2019-06-11 12:27:20 +02:00
|
|
|
g_free(s->info.x_dirty_bitmap);
|
|
|
|
g_free(s->info.name);
|
2019-06-11 12:27:19 +02:00
|
|
|
if (ret < 0) {
|
2020-12-28 16:08:44 +01:00
|
|
|
yank_unregister_function(BLOCKDEV_YANK_INSTANCE(bs->node_name),
|
|
|
|
nbd_yank, bs);
|
|
|
|
object_unref(OBJECT(s->sioc));
|
2020-07-27 20:47:49 +02:00
|
|
|
s->sioc = NULL;
|
2019-06-11 12:27:19 +02:00
|
|
|
return ret;
|
|
|
|
}
|
2020-10-27 06:05:55 +01:00
|
|
|
if (s->x_dirty_bitmap) {
|
|
|
|
if (!s->info.base_allocation) {
|
|
|
|
error_setg(errp, "requested x-dirty-bitmap %s not found",
|
|
|
|
s->x_dirty_bitmap);
|
|
|
|
ret = -EINVAL;
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
if (strcmp(s->x_dirty_bitmap, "qemu:allocation-depth") == 0) {
|
|
|
|
s->alloc_depth = true;
|
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
2019-06-11 12:27:20 +02:00
|
|
|
if (s->info.flags & NBD_FLAG_READ_ONLY) {
|
2019-06-11 12:27:19 +02:00
|
|
|
ret = bdrv_apply_auto_read_only(bs, "NBD export is read-only", errp);
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
}
|
2019-06-11 12:27:20 +02:00
|
|
|
if (s->info.flags & NBD_FLAG_SEND_FUA) {
|
2019-06-11 12:27:19 +02:00
|
|
|
bs->supported_write_flags = BDRV_REQ_FUA;
|
|
|
|
bs->supported_zero_flags |= BDRV_REQ_FUA;
|
|
|
|
}
|
2019-06-11 12:27:20 +02:00
|
|
|
if (s->info.flags & NBD_FLAG_SEND_WRITE_ZEROES) {
|
2019-06-11 12:27:19 +02:00
|
|
|
bs->supported_zero_flags |= BDRV_REQ_MAY_UNMAP;
|
2019-08-23 16:37:24 +02:00
|
|
|
if (s->info.flags & NBD_FLAG_SEND_FAST_ZERO) {
|
|
|
|
bs->supported_zero_flags |= BDRV_REQ_NO_FALLBACK;
|
|
|
|
}
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
if (!s->ioc) {
|
2020-12-28 16:08:44 +01:00
|
|
|
s->ioc = QIO_CHANNEL(s->sioc);
|
2019-06-11 12:27:20 +02:00
|
|
|
object_ref(OBJECT(s->ioc));
|
2019-06-11 12:27:19 +02:00
|
|
|
}
|
|
|
|
|
2020-07-27 20:47:47 +02:00
|
|
|
trace_nbd_client_handshake_success(s->export);
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
fail:
|
|
|
|
/*
|
2019-06-18 13:43:21 +02:00
|
|
|
* We have connected, but must fail for other reasons.
|
|
|
|
* Send NBD_CMD_DISC as a courtesy to the server.
|
2019-06-11 12:27:19 +02:00
|
|
|
*/
|
|
|
|
{
|
|
|
|
NBDRequest request = { .type = NBD_CMD_DISC };
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
nbd_send_request(s->ioc ?: QIO_CHANNEL(s->sioc), &request);
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
yank_unregister_function(BLOCKDEV_YANK_INSTANCE(bs->node_name),
|
|
|
|
nbd_yank, bs);
|
|
|
|
object_unref(OBJECT(s->sioc));
|
2020-07-27 20:47:49 +02:00
|
|
|
s->sioc = NULL;
|
2021-06-10 12:07:38 +02:00
|
|
|
object_unref(OBJECT(s->ioc));
|
|
|
|
s->ioc = NULL;
|
2019-06-11 12:27:19 +02:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-06-18 13:43:24 +02:00
|
|
|
/*
|
|
|
|
* Parse nbd_open options
|
|
|
|
*/
|
2019-06-11 12:27:19 +02:00
|
|
|
|
2013-03-07 16:15:11 +01:00
|
|
|
static int nbd_parse_uri(const char *filename, QDict *options)
|
2012-11-04 13:04:24 +01:00
|
|
|
{
|
|
|
|
URI *uri;
|
|
|
|
const char *p;
|
|
|
|
QueryParams *qp = NULL;
|
|
|
|
int ret = 0;
|
2013-03-07 16:15:11 +01:00
|
|
|
bool is_unix;
|
2012-11-04 13:04:24 +01:00
|
|
|
|
|
|
|
uri = uri_parse(filename);
|
|
|
|
if (!uri) {
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* transport */
|
2017-06-13 22:57:26 +02:00
|
|
|
if (!g_strcmp0(uri->scheme, "nbd")) {
|
2013-03-07 16:15:11 +01:00
|
|
|
is_unix = false;
|
2017-06-13 22:57:26 +02:00
|
|
|
} else if (!g_strcmp0(uri->scheme, "nbd+tcp")) {
|
2013-03-07 16:15:11 +01:00
|
|
|
is_unix = false;
|
2017-06-13 22:57:26 +02:00
|
|
|
} else if (!g_strcmp0(uri->scheme, "nbd+unix")) {
|
2013-03-07 16:15:11 +01:00
|
|
|
is_unix = true;
|
2012-11-04 13:04:24 +01:00
|
|
|
} else {
|
|
|
|
ret = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2020-02-12 03:31:01 +01:00
|
|
|
p = uri->path ? uri->path : "";
|
|
|
|
if (p[0] == '/') {
|
|
|
|
p++;
|
|
|
|
}
|
2012-11-04 13:04:24 +01:00
|
|
|
if (p[0]) {
|
2017-04-27 23:58:17 +02:00
|
|
|
qdict_put_str(options, "export", p);
|
2012-11-04 13:04:24 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
qp = query_params_parse(uri->query);
|
2013-03-07 16:15:11 +01:00
|
|
|
if (qp->n > 1 || (is_unix && !qp->n) || (!is_unix && qp->n)) {
|
2012-11-04 13:04:24 +01:00
|
|
|
ret = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2013-03-07 16:15:11 +01:00
|
|
|
if (is_unix) {
|
2012-11-04 13:04:24 +01:00
|
|
|
/* nbd+unix:///export?socket=path */
|
|
|
|
if (uri->server || uri->port || strcmp(qp->p[0].name, "socket")) {
|
|
|
|
ret = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
2017-04-27 23:58:17 +02:00
|
|
|
qdict_put_str(options, "server.type", "unix");
|
|
|
|
qdict_put_str(options, "server.path", qp->p[0].value);
|
2012-11-04 13:04:24 +01:00
|
|
|
} else {
|
2013-06-03 17:54:56 +02:00
|
|
|
QString *host;
|
2016-10-25 15:11:35 +02:00
|
|
|
char *port_str;
|
|
|
|
|
2013-03-18 16:56:05 +01:00
|
|
|
/* nbd[+tcp]://host[:port]/export */
|
2012-11-04 13:04:24 +01:00
|
|
|
if (!uri->server) {
|
|
|
|
ret = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
2013-03-15 11:55:29 +01:00
|
|
|
|
2013-06-03 17:54:56 +02:00
|
|
|
/* strip braces from literal IPv6 address */
|
|
|
|
if (uri->server[0] == '[') {
|
|
|
|
host = qstring_from_substr(uri->server, 1,
|
2018-07-27 08:22:04 +02:00
|
|
|
strlen(uri->server) - 1);
|
2013-06-03 17:54:56 +02:00
|
|
|
} else {
|
|
|
|
host = qstring_from_str(uri->server);
|
|
|
|
}
|
|
|
|
|
2017-04-27 23:58:17 +02:00
|
|
|
qdict_put_str(options, "server.type", "inet");
|
nbd: Tidy up blockdev-add interface
SocketAddress is a simple union, and simple unions are awkward: they
have their variant members wrapped in a "data" object on the wire, and
require additional indirections in C. I intend to limit its use to
existing external interfaces, and convert all internal interfaces to
SocketAddressFlat.
BlockdevOptionsNbd is an external interface using SocketAddress. We
already use SocketAddressFlat elsewhere in blockdev-add. Replace it
by SocketAddressFlat while we can (it's new in 2.9) for simplicity and
consistency. For example,
{ "execute": "blockdev-add",
"arguments": { "node-name": "foo", "driver": "nbd",
"server": { "type": "inet",
"data": { "host": "localhost",
"port": "12345" } } } }
becomes
{ "execute": "blockdev-add",
"arguments": { "node-name": "foo", "driver": "nbd",
"server": { "type": "inet",
"host": "localhost", "port": "12345" } } }
Since the internal interfaces still take SocketAddress, this requires
conversion function socket_address_crumple(). It'll go away when I
update the interfaces.
Unfortunately, SocketAddress is also visible in -drive since 2.8:
-drive if=none,driver=nbd,server.type=inet,server.data.host=127.0.0.1,server.data.port=12345
Nobody should be using it, as it's fairly new and has never been
documented, so adding still more compatibility gunk to keep it working
isn't worth the trouble. You now have to use
-drive if=none,driver=nbd,server.type=inet,server.host=127.0.0.1,server.port=12345
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1490895797-29094-9-git-send-email-armbru@redhat.com
[mreitz: Change iotest 147 accordingly]
Because of this interface change, iotest 147 has to be adapted.
Unfortunately, we cannot just flatten all of the addresses because
nbd-server-start still takes a plain SocketAddress. Therefore, we need
both and this is most easily achieved by writing the SocketAddress into
the code and flattening it where necessary.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20170330221243.17333-1-mreitz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
2017-03-30 19:43:16 +02:00
|
|
|
qdict_put(options, "server.host", host);
|
2016-10-25 15:11:35 +02:00
|
|
|
|
|
|
|
port_str = g_strdup_printf("%d", uri->port ?: NBD_DEFAULT_PORT);
|
2017-04-27 23:58:17 +02:00
|
|
|
qdict_put_str(options, "server.port", port_str);
|
2016-10-25 15:11:35 +02:00
|
|
|
g_free(port_str);
|
2012-11-04 13:04:24 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
|
|
|
if (qp) {
|
|
|
|
query_params_free(qp);
|
|
|
|
}
|
|
|
|
uri_free(uri);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2016-10-25 15:11:33 +02:00
|
|
|
static bool nbd_has_filename_options_conflict(QDict *options, Error **errp)
|
|
|
|
{
|
|
|
|
const QDictEntry *e;
|
|
|
|
|
|
|
|
for (e = qdict_first(options); e; e = qdict_next(options, e)) {
|
|
|
|
if (!strcmp(e->key, "host") ||
|
|
|
|
!strcmp(e->key, "port") ||
|
|
|
|
!strcmp(e->key, "path") ||
|
2016-10-25 15:11:34 +02:00
|
|
|
!strcmp(e->key, "export") ||
|
|
|
|
strstart(e->key, "server.", NULL))
|
2016-10-25 15:11:33 +02:00
|
|
|
{
|
|
|
|
error_setg(errp, "Option '%s' cannot be used with a file name",
|
|
|
|
e->key);
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2013-03-15 18:47:22 +01:00
|
|
|
static void nbd_parse_filename(const char *filename, QDict *options,
|
|
|
|
Error **errp)
|
2008-07-03 15:41:03 +02:00
|
|
|
{
|
2019-08-24 19:28:12 +02:00
|
|
|
g_autofree char *file = NULL;
|
2011-02-22 16:44:54 +01:00
|
|
|
char *export_name;
|
|
|
|
const char *host_spec;
|
2008-07-03 15:41:03 +02:00
|
|
|
const char *unixpath;
|
|
|
|
|
2016-10-25 15:11:33 +02:00
|
|
|
if (nbd_has_filename_options_conflict(options, errp)) {
|
2013-03-20 19:23:23 +01:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2012-11-04 13:04:24 +01:00
|
|
|
if (strstr(filename, "://")) {
|
2013-03-15 18:47:22 +01:00
|
|
|
int ret = nbd_parse_uri(filename, options);
|
|
|
|
if (ret < 0) {
|
|
|
|
error_setg(errp, "No valid URL specified");
|
|
|
|
}
|
|
|
|
return;
|
2012-11-04 13:04:24 +01:00
|
|
|
}
|
|
|
|
|
2011-08-21 05:09:37 +02:00
|
|
|
file = g_strdup(filename);
|
2010-08-25 22:48:33 +02:00
|
|
|
|
2011-02-22 16:44:54 +01:00
|
|
|
export_name = strstr(file, EN_OPTSTR);
|
|
|
|
if (export_name) {
|
|
|
|
if (export_name[strlen(EN_OPTSTR)] == 0) {
|
2019-08-24 19:28:12 +02:00
|
|
|
return;
|
2010-08-25 22:48:33 +02:00
|
|
|
}
|
2011-02-22 16:44:54 +01:00
|
|
|
export_name[0] = 0; /* truncate 'file' */
|
|
|
|
export_name += strlen(EN_OPTSTR);
|
2013-03-07 16:15:11 +01:00
|
|
|
|
2017-04-27 23:58:17 +02:00
|
|
|
qdict_put_str(options, "export", export_name);
|
2010-08-25 22:48:33 +02:00
|
|
|
}
|
|
|
|
|
2011-02-22 16:44:54 +01:00
|
|
|
/* extract the host_spec - fail if it's not nbd:... */
|
|
|
|
if (!strstart(file, "nbd:", &host_spec)) {
|
2013-03-15 18:47:22 +01:00
|
|
|
error_setg(errp, "File name string for NBD must start with 'nbd:'");
|
2019-08-24 19:28:12 +02:00
|
|
|
return;
|
2010-08-25 22:48:33 +02:00
|
|
|
}
|
2008-07-03 15:41:03 +02:00
|
|
|
|
2013-03-07 16:15:11 +01:00
|
|
|
if (!*host_spec) {
|
2019-08-24 19:28:12 +02:00
|
|
|
return;
|
2013-03-07 16:15:11 +01:00
|
|
|
}
|
|
|
|
|
2011-02-22 16:44:54 +01:00
|
|
|
/* are we a UNIX or TCP socket? */
|
|
|
|
if (strstart(host_spec, "unix:", &unixpath)) {
|
2017-04-27 23:58:17 +02:00
|
|
|
qdict_put_str(options, "server.type", "unix");
|
|
|
|
qdict_put_str(options, "server.path", unixpath);
|
2008-07-03 15:41:03 +02:00
|
|
|
} else {
|
2017-04-26 09:36:37 +02:00
|
|
|
InetSocketAddress *addr = g_new(InetSocketAddress, 1);
|
2013-03-07 16:15:11 +01:00
|
|
|
|
2017-04-26 09:36:37 +02:00
|
|
|
if (inet_parse(addr, host_spec, errp)) {
|
|
|
|
goto out_inet;
|
2013-03-15 11:55:29 +01:00
|
|
|
}
|
2008-07-03 15:41:03 +02:00
|
|
|
|
2017-04-27 23:58:17 +02:00
|
|
|
qdict_put_str(options, "server.type", "inet");
|
|
|
|
qdict_put_str(options, "server.host", addr->host);
|
|
|
|
qdict_put_str(options, "server.port", addr->port);
|
2017-04-26 09:36:37 +02:00
|
|
|
out_inet:
|
2013-03-07 16:15:11 +01:00
|
|
|
qapi_free_InetSocketAddress(addr);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-10-25 15:11:34 +02:00
|
|
|
static bool nbd_process_legacy_socket_options(QDict *output_options,
|
|
|
|
QemuOpts *legacy_opts,
|
|
|
|
Error **errp)
|
2013-03-07 16:15:11 +01:00
|
|
|
{
|
2016-10-25 15:11:34 +02:00
|
|
|
const char *path = qemu_opt_get(legacy_opts, "path");
|
|
|
|
const char *host = qemu_opt_get(legacy_opts, "host");
|
|
|
|
const char *port = qemu_opt_get(legacy_opts, "port");
|
|
|
|
const QDictEntry *e;
|
2013-03-07 16:15:11 +01:00
|
|
|
|
2016-10-25 15:11:34 +02:00
|
|
|
if (!path && !host && !port) {
|
|
|
|
return true;
|
|
|
|
}
|
2016-08-15 15:29:26 +02:00
|
|
|
|
2016-10-25 15:11:34 +02:00
|
|
|
for (e = qdict_first(output_options); e; e = qdict_next(output_options, e))
|
|
|
|
{
|
|
|
|
if (strstart(e->key, "server.", NULL)) {
|
|
|
|
error_setg(errp, "Cannot use 'server' and path/host/port at the "
|
|
|
|
"same time");
|
|
|
|
return false;
|
2013-03-20 19:23:23 +01:00
|
|
|
}
|
2011-02-22 16:44:54 +01:00
|
|
|
}
|
2016-10-25 15:11:34 +02:00
|
|
|
|
|
|
|
if (path && host) {
|
|
|
|
error_setg(errp, "path and host may not be used at the same time");
|
|
|
|
return false;
|
|
|
|
} else if (path) {
|
|
|
|
if (port) {
|
|
|
|
error_setg(errp, "port may not be used without host");
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2017-04-27 23:58:17 +02:00
|
|
|
qdict_put_str(output_options, "server.type", "unix");
|
|
|
|
qdict_put_str(output_options, "server.path", path);
|
2016-10-25 15:11:34 +02:00
|
|
|
} else if (host) {
|
2017-04-27 23:58:17 +02:00
|
|
|
qdict_put_str(output_options, "server.type", "inet");
|
|
|
|
qdict_put_str(output_options, "server.host", host);
|
|
|
|
qdict_put_str(output_options, "server.port",
|
|
|
|
port ?: stringify(NBD_DEFAULT_PORT));
|
2016-10-25 15:11:30 +02:00
|
|
|
}
|
2013-03-07 16:15:11 +01:00
|
|
|
|
2016-10-25 15:11:34 +02:00
|
|
|
return true;
|
|
|
|
}
|
2013-03-07 16:15:11 +01:00
|
|
|
|
2017-04-26 09:36:40 +02:00
|
|
|
static SocketAddress *nbd_config(BDRVNBDState *s, QDict *options,
|
|
|
|
Error **errp)
|
2016-10-25 15:11:34 +02:00
|
|
|
{
|
2017-04-26 09:36:40 +02:00
|
|
|
SocketAddress *saddr = NULL;
|
2016-10-25 15:11:34 +02:00
|
|
|
QDict *addr = NULL;
|
|
|
|
Visitor *iv = NULL;
|
|
|
|
|
|
|
|
qdict_extract_subqdict(options, &addr, "server.");
|
|
|
|
if (!qdict_size(addr)) {
|
|
|
|
error_setg(errp, "NBD server address missing");
|
|
|
|
goto done;
|
2013-03-07 16:15:11 +01:00
|
|
|
}
|
|
|
|
|
2018-06-14 21:14:33 +02:00
|
|
|
iv = qobject_input_visitor_new_flat_confused(addr, errp);
|
|
|
|
if (!iv) {
|
2016-10-25 15:11:34 +02:00
|
|
|
goto done;
|
|
|
|
}
|
2013-03-18 16:56:05 +01:00
|
|
|
|
2020-07-07 18:06:03 +02:00
|
|
|
if (!visit_type_SocketAddress(iv, NULL, &saddr, errp)) {
|
2016-10-25 15:11:34 +02:00
|
|
|
goto done;
|
|
|
|
}
|
2015-09-16 15:52:22 +02:00
|
|
|
|
2021-06-10 12:07:36 +02:00
|
|
|
if (socket_address_parse_named_fd(saddr, errp) < 0) {
|
|
|
|
qapi_free_SocketAddress(saddr);
|
|
|
|
saddr = NULL;
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
2016-10-25 15:11:34 +02:00
|
|
|
done:
|
2018-04-19 17:01:43 +02:00
|
|
|
qobject_unref(addr);
|
2016-10-25 15:11:34 +02:00
|
|
|
visit_free(iv);
|
2015-09-16 15:52:22 +02:00
|
|
|
return saddr;
|
2011-02-22 16:44:54 +01:00
|
|
|
}
|
2010-08-25 22:48:33 +02:00
|
|
|
|
2016-02-10 19:41:12 +01:00
|
|
|
static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, Error **errp)
|
|
|
|
{
|
|
|
|
Object *obj;
|
|
|
|
QCryptoTLSCreds *creds;
|
|
|
|
|
|
|
|
obj = object_resolve_path_component(
|
|
|
|
object_get_objects_root(), id);
|
|
|
|
if (!obj) {
|
|
|
|
error_setg(errp, "No TLS credentials with id '%s'",
|
|
|
|
id);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
creds = (QCryptoTLSCreds *)
|
|
|
|
object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS);
|
|
|
|
if (!creds) {
|
|
|
|
error_setg(errp, "Object with id '%s' is not TLS credentials",
|
|
|
|
id);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_CLIENT) {
|
|
|
|
error_setg(errp,
|
|
|
|
"Expecting TLS credentials with a client endpoint");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
object_ref(obj);
|
|
|
|
return creds;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2016-08-15 15:29:24 +02:00
|
|
|
static QemuOptsList nbd_runtime_opts = {
|
|
|
|
.name = "nbd",
|
|
|
|
.head = QTAILQ_HEAD_INITIALIZER(nbd_runtime_opts.head),
|
|
|
|
.desc = {
|
|
|
|
{
|
|
|
|
.name = "host",
|
|
|
|
.type = QEMU_OPT_STRING,
|
|
|
|
.help = "TCP host to connect to",
|
|
|
|
},
|
|
|
|
{
|
|
|
|
.name = "port",
|
|
|
|
.type = QEMU_OPT_STRING,
|
|
|
|
.help = "TCP port to connect to",
|
|
|
|
},
|
|
|
|
{
|
|
|
|
.name = "path",
|
|
|
|
.type = QEMU_OPT_STRING,
|
|
|
|
.help = "Unix socket path to connect to",
|
|
|
|
},
|
|
|
|
{
|
|
|
|
.name = "export",
|
|
|
|
.type = QEMU_OPT_STRING,
|
|
|
|
.help = "Name of the NBD export to open",
|
|
|
|
},
|
|
|
|
{
|
|
|
|
.name = "tls-creds",
|
|
|
|
.type = QEMU_OPT_STRING,
|
|
|
|
.help = "ID of the TLS credentials to use",
|
|
|
|
},
|
2018-07-02 21:14:57 +02:00
|
|
|
{
|
|
|
|
.name = "x-dirty-bitmap",
|
|
|
|
.type = QEMU_OPT_STRING,
|
|
|
|
.help = "experimental: expose named dirty bitmap in place of "
|
|
|
|
"block status",
|
|
|
|
},
|
2019-06-18 13:43:23 +02:00
|
|
|
{
|
|
|
|
.name = "reconnect-delay",
|
|
|
|
.type = QEMU_OPT_NUMBER,
|
|
|
|
.help = "On an unexpected disconnect, the nbd client tries to "
|
|
|
|
"connect again until succeeding or encountering a serious "
|
|
|
|
"error. During the first @reconnect-delay seconds, all "
|
|
|
|
"requests are paused and will be rerun on a successful "
|
|
|
|
"reconnect. After that time, any delayed requests and all "
|
|
|
|
"future requests before a successful reconnect will "
|
|
|
|
"immediately fail. Default 0",
|
|
|
|
},
|
block/nbd: fix segmentation fault when .desc is not null-terminated
The find_desc_by_name() from util/qemu-option.c relies on the .name not being
NULL to call strcmp(). This check becomes unsafe when the list is not
NULL-terminated, which is the case of nbd_runtime_opts in block/nbd.c, and can
result in segmentation fault when strcmp() tries to access an invalid memory:
#0 0x00007fff8c75f7d4 in __strcmp_power9 () from /lib64/libc.so.6
#1 0x00000000102d3ec8 in find_desc_by_name (desc=0x1036d6f0, name=0x28e46670 "server.path") at util/qemu-option.c:166
#2 0x00000000102d93e0 in qemu_opts_absorb_qdict (opts=0x28e47a80, qdict=0x28e469a0, errp=0x7fffec247c98) at util/qemu-option.c:1026
#3 0x000000001012a2e4 in nbd_open (bs=0x28e42290, options=0x28e469a0, flags=24578, errp=0x7fffec247d80) at block/nbd.c:406
#4 0x00000000100144e8 in bdrv_open_driver (bs=0x28e42290, drv=0x1036e070 <bdrv_nbd_unix>, node_name=0x0, options=0x28e469a0, open_flags=24578, errp=0x7fffec247f50) at block.c:1135
#5 0x0000000010015b04 in bdrv_open_common (bs=0x28e42290, file=0x0, options=0x28e469a0, errp=0x7fffec247f50) at block.c:1395
>From gdb, the desc[i].name was not NULL and resulted in strcmp() accessing an
invalid memory:
>>> p desc[5]
$8 = {
name = 0x1037f098 "R27A",
type = 1561964883,
help = 0xc0bbb23e <error: Cannot access memory at address 0xc0bbb23e>,
def_value_str = 0x2 <error: Cannot access memory at address 0x2>
}
>>> p desc[6]
$9 = {
name = 0x103dac78 <__gcov0.do_qemu_init_bdrv_nbd_init> "\001",
type = 272101528,
help = 0x29ec0b754403e31f <error: Cannot access memory at address 0x29ec0b754403e31f>,
def_value_str = 0x81f343b9 <error: Cannot access memory at address 0x81f343b9>
}
This patch fixes the segmentation fault in strcmp() by adding a NULL element at
the end of nbd_runtime_opts.desc list, which is the common practice to most of
other structs like runtime_opts in block/null.c. Thus, the desc[i].name != NULL
check becomes safe because it will not evaluate to true when .desc list reached
its end.
Reported-by: R. Nageswara Sastry <nasastry@in.ibm.com>
Buglink: https://bugs.launchpad.net/qemu/+bug/1727259
Signed-off-by: Murilo Opsfelder Araujo <muriloo@linux.vnet.ibm.com>
Message-Id: <20180105133241.14141-2-muriloo@linux.vnet.ibm.com>
CC: qemu-stable@nongnu.org
Fixes: 7ccc44fd7d1dfa62c4d6f3a680df809d6e7068ce
Signed-off-by: Eric Blake <eblake@redhat.com>
2018-01-05 14:32:41 +01:00
|
|
|
{ /* end of list */ }
|
2016-08-15 15:29:24 +02:00
|
|
|
},
|
|
|
|
};
|
|
|
|
|
2019-06-18 13:43:24 +02:00
|
|
|
static int nbd_process_options(BlockDriverState *bs, QDict *options,
|
|
|
|
Error **errp)
|
2011-02-22 16:44:54 +01:00
|
|
|
{
|
|
|
|
BDRVNBDState *s = bs->opaque;
|
2019-06-18 13:43:24 +02:00
|
|
|
QemuOpts *opts;
|
2016-02-10 19:41:12 +01:00
|
|
|
int ret = -EINVAL;
|
2011-09-08 14:28:59 +02:00
|
|
|
|
2016-08-15 15:29:24 +02:00
|
|
|
opts = qemu_opts_create(&nbd_runtime_opts, NULL, 0, &error_abort);
|
2020-07-07 18:06:03 +02:00
|
|
|
if (!qemu_opts_absorb_qdict(opts, options, errp)) {
|
2016-08-15 15:29:24 +02:00
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
2017-04-26 09:36:40 +02:00
|
|
|
/* Translate @host, @port, and @path to a SocketAddress */
|
2016-10-25 15:11:34 +02:00
|
|
|
if (!nbd_process_legacy_socket_options(options, opts, errp)) {
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
2011-02-22 16:44:54 +01:00
|
|
|
/* Pop the config into our state object. Exit if invalid. */
|
2016-10-25 15:11:34 +02:00
|
|
|
s->saddr = nbd_config(s, options, errp);
|
|
|
|
if (!s->saddr) {
|
2016-02-10 19:41:12 +01:00
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
2016-10-25 15:11:34 +02:00
|
|
|
s->export = g_strdup(qemu_opt_get(opts, "export"));
|
2019-11-14 03:46:34 +01:00
|
|
|
if (s->export && strlen(s->export) > NBD_MAX_STRING_SIZE) {
|
|
|
|
error_setg(errp, "export name too long to send to server");
|
|
|
|
goto error;
|
|
|
|
}
|
2016-10-25 15:11:34 +02:00
|
|
|
|
2016-08-15 15:29:26 +02:00
|
|
|
s->tlscredsid = g_strdup(qemu_opt_get(opts, "tls-creds"));
|
|
|
|
if (s->tlscredsid) {
|
2019-06-18 13:43:24 +02:00
|
|
|
s->tlscreds = nbd_get_tls_creds(s->tlscredsid, errp);
|
|
|
|
if (!s->tlscreds) {
|
2016-02-10 19:41:12 +01:00
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
2017-03-30 19:43:09 +02:00
|
|
|
/* TODO SOCKET_ADDRESS_KIND_FD where fd has AF_INET or AF_INET6 */
|
2017-04-26 09:36:40 +02:00
|
|
|
if (s->saddr->type != SOCKET_ADDRESS_TYPE_INET) {
|
2016-02-10 19:41:12 +01:00
|
|
|
error_setg(errp, "TLS only supported over IP sockets");
|
|
|
|
goto error;
|
|
|
|
}
|
2019-06-18 13:43:24 +02:00
|
|
|
s->hostname = s->saddr->u.inet.host;
|
2011-02-22 16:44:54 +01:00
|
|
|
}
|
|
|
|
|
2019-06-18 13:43:24 +02:00
|
|
|
s->x_dirty_bitmap = g_strdup(qemu_opt_get(opts, "x-dirty-bitmap"));
|
2019-11-14 03:46:34 +01:00
|
|
|
if (s->x_dirty_bitmap && strlen(s->x_dirty_bitmap) > NBD_MAX_STRING_SIZE) {
|
|
|
|
error_setg(errp, "x-dirty-bitmap query too long to send to server");
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
2019-06-18 13:43:24 +02:00
|
|
|
s->reconnect_delay = qemu_opt_get_number(opts, "reconnect-delay", 0);
|
|
|
|
|
|
|
|
ret = 0;
|
2019-02-01 14:01:34 +01:00
|
|
|
|
2016-02-10 19:41:12 +01:00
|
|
|
error:
|
2016-08-15 15:29:24 +02:00
|
|
|
qemu_opts_del(opts);
|
2016-02-10 19:41:12 +01:00
|
|
|
return ret;
|
2011-10-20 13:16:23 +02:00
|
|
|
}
|
|
|
|
|
2019-06-18 13:43:24 +02:00
|
|
|
static int nbd_open(BlockDriverState *bs, QDict *options, int flags,
|
|
|
|
Error **errp)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
|
|
|
|
|
|
|
s->bs = bs;
|
|
|
|
qemu_co_mutex_init(&s->send_mutex);
|
|
|
|
qemu_co_queue_init(&s->free_sema);
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
if (!yank_register_instance(BLOCKDEV_YANK_INSTANCE(bs->node_name), errp)) {
|
|
|
|
return -EEXIST;
|
|
|
|
}
|
|
|
|
|
2021-06-10 12:07:33 +02:00
|
|
|
ret = nbd_process_options(bs, options, errp);
|
|
|
|
if (ret < 0) {
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
|
2021-06-10 12:07:37 +02:00
|
|
|
nbd_init_connect_thread(s);
|
|
|
|
|
2020-07-27 20:47:47 +02:00
|
|
|
/*
|
|
|
|
* establish TCP connection, return error if it fails
|
|
|
|
* TODO: Configurable retry-until-timeout behaviour.
|
|
|
|
*/
|
2020-12-28 16:08:44 +01:00
|
|
|
if (nbd_establish_connection(bs, s->saddr, errp) < 0) {
|
2021-06-10 12:07:33 +02:00
|
|
|
ret = -ECONNREFUSED;
|
|
|
|
goto fail;
|
2020-07-27 20:47:47 +02:00
|
|
|
}
|
|
|
|
|
2020-12-28 16:08:44 +01:00
|
|
|
ret = nbd_client_handshake(bs, errp);
|
2019-06-18 13:43:24 +02:00
|
|
|
if (ret < 0) {
|
2021-06-10 12:07:33 +02:00
|
|
|
goto fail;
|
2019-06-18 13:43:24 +02:00
|
|
|
}
|
|
|
|
/* successfully connected */
|
|
|
|
s->state = NBD_CLIENT_CONNECTED;
|
|
|
|
|
|
|
|
s->connection_co = qemu_coroutine_create(nbd_connection_entry, s);
|
|
|
|
bdrv_inc_in_flight(bs);
|
|
|
|
aio_co_schedule(bdrv_get_aio_context(bs), s->connection_co);
|
|
|
|
|
|
|
|
return 0;
|
2021-06-10 12:07:33 +02:00
|
|
|
|
|
|
|
fail:
|
|
|
|
nbd_clear_bdrvstate(bs);
|
|
|
|
return ret;
|
2019-06-18 13:43:24 +02:00
|
|
|
}
|
|
|
|
|
2011-10-21 13:17:14 +02:00
|
|
|
static int nbd_co_flush(BlockDriverState *bs)
|
|
|
|
{
|
2015-02-06 22:06:16 +01:00
|
|
|
return nbd_client_co_flush(bs);
|
2011-10-21 13:17:14 +02:00
|
|
|
}
|
|
|
|
|
2015-02-06 12:24:43 +01:00
|
|
|
static void nbd_refresh_limits(BlockDriverState *bs, Error **errp)
|
|
|
|
{
|
2019-06-11 12:27:20 +02:00
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
nbd: Honor server's advertised minimum block size
Commit 79ba8c98 (v2.7) changed the setting of request_alignment
to occur only during bdrv_refresh_limits(), rather than at at
bdrv_open() time; but at the time, NBD was unaffected, because
it still used sector-based callbacks, so the block layer
defaulted NBD to use 512 request_alignment.
Later, commit 70c4fb26 (also v2.7) changed NBD to use byte-based
callbacks, without setting request_alignment. This resulted in
NBD using request_alignment of 1, which works great when the
server supports it (as is the case for qemu-nbd), but falls apart
miserably if the server requires alignment (but only if qemu
actually sends a sub-sector request; qemu-io can do it, but
most qemu operations still perform on sectors or larger).
Even later, the NBD protocol was updated to document that clients
should learn the server's minimum alignment during NBD_OPT_GO;
and recommended that clients should assume a minimum size of 512
unless the server understands NBD_OPT_GO and replied with a smaller
size. Commit 081dd1fe (v2.10) attempted to do that, by assigning
request_alignment to whatever was learned from the server; but
it has two flaws: the assignment is done during bdrv_open() so
it gets unconditionally wiped out back to 1 during any later
bdrv_refresh_limits(); and the code is not using a default of 512
when the server did not report a minimum size.
Fix these issues by moving the assignment to request_alignment
to the right function, and by using a sane default when the
server does not advertise a minimum size.
CC: qemu-stable@nongnu.org
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20180215032905.27146-1-eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy<vsementsov@virtuozzo.com>
2018-02-15 04:29:05 +01:00
|
|
|
uint32_t min = s->info.min_block;
|
nbd: Implement NBD_INFO_BLOCK_SIZE on client
The upstream NBD Protocol has defined a new extension to allow
the server to advertise block sizes to the client, as well as
a way for the client to inform the server whether it intends to
obey block sizes.
When using the block layer as the client, we will obey block
sizes; but when used as 'qemu-nbd -c' to hand off to the
kernel nbd module as the client, we are still waiting for the
kernel to implement a way for us to learn if it will honor
block sizes (perhaps by an addition to sysfs, rather than an
ioctl), as well as any way to tell the kernel what additional
block sizes to obey (NBD_SET_BLKSIZE appears to be accurate
for the minimum size, but preferred and maximum sizes would
probably be new ioctl()s), so until then, we need to make our
request for block sizes conditional.
When using ioctl(NBD_SET_BLKSIZE) to hand off to the kernel,
use the minimum block size as the sector size if it is larger
than 512, which also has the nice effect of cooperating with
(non-qemu) servers that don't do read-modify-write when
exposing a block device with 4k sectors; it might also allow
us to visit a file larger than 2T on a 32-bit kernel.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170707203049.534-10-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-07-07 22:30:49 +02:00
|
|
|
uint32_t max = MIN_NON_ZERO(NBD_MAX_BUFFER_SIZE, s->info.max_block);
|
|
|
|
|
nbd/client: Lower min_block for block-status, unaligned size
We have a latent bug in our NBD client code, tickled by the brand new
nbdkit 1.11.10 block status support:
$ nbdkit --filter=log --filter=truncate -U - \
data data="1" size=511 truncate=64K logfile=/dev/stdout \
--run 'qemu-img convert $nbd /var/tmp/out'
...
qemu-img: block/io.c:2122: bdrv_co_block_status: Assertion `*pnum && QEMU_IS_ALIGNED(*pnum, align) && align > offset - aligned_offset' failed.
The culprit? Our implementation of .bdrv_co_block_status can return
unaligned block status for any server that operates with a lower
actual alignment than what we tell the block layer in
request_alignment, in violation of the block layer's constraints. To
date, we've been unable to trip the bug, because qemu as NBD server
always advertises block sizing (at which point it is a server bug if
the server sends unaligned status - although qemu 3.1 is such a server
and I've sent separate patches for 4.0 both to get the server to obey
the spec, and to let the client to tolerate server oddities at EOF).
But nbdkit does not (yet) advertise block sizing, and therefore is not
in violation of the spec for returning block status at whatever
boundaries it wants, and those unaligned results can occur anywhere
rather than just at EOF. While we are still wise to avoid sending
sub-sector read/write requests to a server of unknown origin, we MUST
consider that a server telling us block status without an advertised
block size is correct. So, we either have to munge unaligned answers
from the server into aligned ones that we hand back to the block
layer, or we have to tell the block layer about a smaller alignment.
Similarly, if the server advertises an image size that is not
sector-aligned, we might as well assume that the server intends to let
us access those tail bytes, and therefore supports a minimum block
size of 1, regardless of whether the server supports block status
(although we still need more patches to fix the problem that with an
unaligned image, we can send read or block status requests that exceed
EOF to the server). Again, qemu as server cannot trip this problem
(because it rounds images to sector alignment), but nbdkit advertised
unaligned size even before it gained block status support.
Solve both alignment problems at once by using better heuristics on
what alignment to report to the block layer when the server did not
give us something to work with. Note that very few NBD servers
implement block status (to date, only qemu and nbdkit are known to do
so); and as the NBD spec mentioned block sizing constraints prior to
documenting block status, it can be assumed that any future
implementations of block status are aware that they must advertise
block size if they want a minimum size other than 1.
We've had a long history of struggles with picking the right alignment
to use in the block layer, as evidenced by the commit message of
fd8d372d (v2.12) that introduced the current choice of forced 512-byte
alignment.
There is no iotest coverage for this fix, because qemu can't provoke
it, and I didn't want to make test 241 dependent on nbdkit.
Fixes: fd8d372d
Reported-by: Richard W.M. Jones <rjones@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20190329042750.14704-3-eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Tested-by: Richard W.M. Jones <rjones@redhat.com>
2019-03-29 05:27:46 +01:00
|
|
|
/*
|
|
|
|
* If the server did not advertise an alignment:
|
|
|
|
* - a size that is not sector-aligned implies that an alignment
|
|
|
|
* of 1 can be used to access those tail bytes
|
|
|
|
* - advertisement of block status requires an alignment of 1, so
|
|
|
|
* that we don't violate block layer constraints that block
|
|
|
|
* status is always aligned (as we can't control whether the
|
|
|
|
* server will report sub-sector extents, such as a hole at EOF
|
|
|
|
* on an unaligned POSIX file)
|
|
|
|
* - otherwise, assume the server is so old that we are safer avoiding
|
|
|
|
* sub-sector requests
|
|
|
|
*/
|
|
|
|
if (!min) {
|
|
|
|
min = (!QEMU_IS_ALIGNED(s->info.size, BDRV_SECTOR_SIZE) ||
|
|
|
|
s->info.base_allocation) ? 1 : BDRV_SECTOR_SIZE;
|
|
|
|
}
|
|
|
|
|
|
|
|
bs->bl.request_alignment = min;
|
2020-04-01 17:01:08 +02:00
|
|
|
bs->bl.max_pdiscard = QEMU_ALIGN_DOWN(INT_MAX, min);
|
nbd: Implement NBD_INFO_BLOCK_SIZE on client
The upstream NBD Protocol has defined a new extension to allow
the server to advertise block sizes to the client, as well as
a way for the client to inform the server whether it intends to
obey block sizes.
When using the block layer as the client, we will obey block
sizes; but when used as 'qemu-nbd -c' to hand off to the
kernel nbd module as the client, we are still waiting for the
kernel to implement a way for us to learn if it will honor
block sizes (perhaps by an addition to sysfs, rather than an
ioctl), as well as any way to tell the kernel what additional
block sizes to obey (NBD_SET_BLKSIZE appears to be accurate
for the minimum size, but preferred and maximum sizes would
probably be new ioctl()s), so until then, we need to make our
request for block sizes conditional.
When using ioctl(NBD_SET_BLKSIZE) to hand off to the kernel,
use the minimum block size as the sector size if it is larger
than 512, which also has the nice effect of cooperating with
(non-qemu) servers that don't do read-modify-write when
exposing a block device with 4k sectors; it might also allow
us to visit a file larger than 2T on a 32-bit kernel.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20170707203049.534-10-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-07-07 22:30:49 +02:00
|
|
|
bs->bl.max_pwrite_zeroes = max;
|
|
|
|
bs->bl.max_transfer = max;
|
|
|
|
|
|
|
|
if (s->info.opt_block &&
|
|
|
|
s->info.opt_block > bs->bl.opt_transfer) {
|
|
|
|
bs->bl.opt_transfer = s->info.opt_block;
|
|
|
|
}
|
2015-02-06 12:24:43 +01:00
|
|
|
}
|
|
|
|
|
2008-07-03 15:41:03 +02:00
|
|
|
static void nbd_close(BlockDriverState *bs)
|
|
|
|
{
|
2015-02-06 22:06:16 +01:00
|
|
|
nbd_client_close(bs);
|
2021-06-10 12:07:33 +02:00
|
|
|
nbd_clear_bdrvstate(bs);
|
2008-07-03 15:41:03 +02:00
|
|
|
}
|
|
|
|
|
2020-07-27 23:58:43 +02:00
|
|
|
/*
|
|
|
|
* NBD cannot truncate, but if the caller asks to truncate to the same size, or
|
|
|
|
* to a smaller size with exact=false, there is no reason to fail the
|
|
|
|
* operation.
|
|
|
|
*
|
|
|
|
* Preallocation mode is ignored since it does not seems useful to fail when
|
|
|
|
* we never change anything.
|
|
|
|
*/
|
|
|
|
static int coroutine_fn nbd_co_truncate(BlockDriverState *bs, int64_t offset,
|
|
|
|
bool exact, PreallocMode prealloc,
|
|
|
|
BdrvRequestFlags flags, Error **errp)
|
|
|
|
{
|
|
|
|
BDRVNBDState *s = bs->opaque;
|
|
|
|
|
|
|
|
if (offset != s->info.size && exact) {
|
|
|
|
error_setg(errp, "Cannot resize NBD nodes");
|
|
|
|
return -ENOTSUP;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (offset > s->info.size) {
|
|
|
|
error_setg(errp, "Cannot grow NBD nodes");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-07-03 15:41:03 +02:00
|
|
|
static int64_t nbd_getlength(BlockDriverState *bs)
|
|
|
|
{
|
|
|
|
BDRVNBDState *s = bs->opaque;
|
|
|
|
|
2019-06-11 12:27:20 +02:00
|
|
|
return s->info.size;
|
2008-07-03 15:41:03 +02:00
|
|
|
}
|
|
|
|
|
2019-02-01 20:29:28 +01:00
|
|
|
static void nbd_refresh_filename(BlockDriverState *bs)
|
2014-07-18 20:24:59 +02:00
|
|
|
{
|
2016-08-15 15:29:26 +02:00
|
|
|
BDRVNBDState *s = bs->opaque;
|
2016-10-25 15:11:34 +02:00
|
|
|
const char *host = NULL, *port = NULL, *path = NULL;
|
2020-06-08 20:26:38 +02:00
|
|
|
size_t len = 0;
|
2016-10-25 15:11:34 +02:00
|
|
|
|
2017-04-26 09:36:40 +02:00
|
|
|
if (s->saddr->type == SOCKET_ADDRESS_TYPE_INET) {
|
nbd: Tidy up blockdev-add interface
SocketAddress is a simple union, and simple unions are awkward: they
have their variant members wrapped in a "data" object on the wire, and
require additional indirections in C. I intend to limit its use to
existing external interfaces, and convert all internal interfaces to
SocketAddressFlat.
BlockdevOptionsNbd is an external interface using SocketAddress. We
already use SocketAddressFlat elsewhere in blockdev-add. Replace it
by SocketAddressFlat while we can (it's new in 2.9) for simplicity and
consistency. For example,
{ "execute": "blockdev-add",
"arguments": { "node-name": "foo", "driver": "nbd",
"server": { "type": "inet",
"data": { "host": "localhost",
"port": "12345" } } } }
becomes
{ "execute": "blockdev-add",
"arguments": { "node-name": "foo", "driver": "nbd",
"server": { "type": "inet",
"host": "localhost", "port": "12345" } } }
Since the internal interfaces still take SocketAddress, this requires
conversion function socket_address_crumple(). It'll go away when I
update the interfaces.
Unfortunately, SocketAddress is also visible in -drive since 2.8:
-drive if=none,driver=nbd,server.type=inet,server.data.host=127.0.0.1,server.data.port=12345
Nobody should be using it, as it's fairly new and has never been
documented, so adding still more compatibility gunk to keep it working
isn't worth the trouble. You now have to use
-drive if=none,driver=nbd,server.type=inet,server.host=127.0.0.1,server.port=12345
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1490895797-29094-9-git-send-email-armbru@redhat.com
[mreitz: Change iotest 147 accordingly]
Because of this interface change, iotest 147 has to be adapted.
Unfortunately, we cannot just flatten all of the addresses because
nbd-server-start still takes a plain SocketAddress. Therefore, we need
both and this is most easily achieved by writing the SocketAddress into
the code and flattening it where necessary.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20170330221243.17333-1-mreitz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
2017-03-30 19:43:16 +02:00
|
|
|
const InetSocketAddress *inet = &s->saddr->u.inet;
|
2016-10-25 15:11:34 +02:00
|
|
|
if (!inet->has_ipv4 && !inet->has_ipv6 && !inet->has_to) {
|
|
|
|
host = inet->host;
|
|
|
|
port = inet->port;
|
|
|
|
}
|
2017-04-26 09:36:40 +02:00
|
|
|
} else if (s->saddr->type == SOCKET_ADDRESS_TYPE_UNIX) {
|
nbd: Tidy up blockdev-add interface
SocketAddress is a simple union, and simple unions are awkward: they
have their variant members wrapped in a "data" object on the wire, and
require additional indirections in C. I intend to limit its use to
existing external interfaces, and convert all internal interfaces to
SocketAddressFlat.
BlockdevOptionsNbd is an external interface using SocketAddress. We
already use SocketAddressFlat elsewhere in blockdev-add. Replace it
by SocketAddressFlat while we can (it's new in 2.9) for simplicity and
consistency. For example,
{ "execute": "blockdev-add",
"arguments": { "node-name": "foo", "driver": "nbd",
"server": { "type": "inet",
"data": { "host": "localhost",
"port": "12345" } } } }
becomes
{ "execute": "blockdev-add",
"arguments": { "node-name": "foo", "driver": "nbd",
"server": { "type": "inet",
"host": "localhost", "port": "12345" } } }
Since the internal interfaces still take SocketAddress, this requires
conversion function socket_address_crumple(). It'll go away when I
update the interfaces.
Unfortunately, SocketAddress is also visible in -drive since 2.8:
-drive if=none,driver=nbd,server.type=inet,server.data.host=127.0.0.1,server.data.port=12345
Nobody should be using it, as it's fairly new and has never been
documented, so adding still more compatibility gunk to keep it working
isn't worth the trouble. You now have to use
-drive if=none,driver=nbd,server.type=inet,server.host=127.0.0.1,server.port=12345
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1490895797-29094-9-git-send-email-armbru@redhat.com
[mreitz: Change iotest 147 accordingly]
Because of this interface change, iotest 147 has to be adapted.
Unfortunately, we cannot just flatten all of the addresses because
nbd-server-start still takes a plain SocketAddress. Therefore, we need
both and this is most easily achieved by writing the SocketAddress into
the code and flattening it where necessary.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20170330221243.17333-1-mreitz@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
2017-03-30 19:43:16 +02:00
|
|
|
path = s->saddr->u.q_unix.path;
|
|
|
|
} /* else can't represent as pseudo-filename */
|
2014-07-18 20:24:59 +02:00
|
|
|
|
2016-10-25 15:11:34 +02:00
|
|
|
if (path && s->export) {
|
2020-06-08 20:26:38 +02:00
|
|
|
len = snprintf(bs->exact_filename, sizeof(bs->exact_filename),
|
|
|
|
"nbd+unix:///%s?socket=%s", s->export, path);
|
2016-10-25 15:11:34 +02:00
|
|
|
} else if (path && !s->export) {
|
2020-06-08 20:26:38 +02:00
|
|
|
len = snprintf(bs->exact_filename, sizeof(bs->exact_filename),
|
|
|
|
"nbd+unix://?socket=%s", path);
|
2016-10-25 15:11:34 +02:00
|
|
|
} else if (host && s->export) {
|
2020-06-08 20:26:38 +02:00
|
|
|
len = snprintf(bs->exact_filename, sizeof(bs->exact_filename),
|
|
|
|
"nbd://%s:%s/%s", host, port, s->export);
|
2016-10-25 15:11:34 +02:00
|
|
|
} else if (host && !s->export) {
|
2020-06-08 20:26:38 +02:00
|
|
|
len = snprintf(bs->exact_filename, sizeof(bs->exact_filename),
|
|
|
|
"nbd://%s:%s", host, port);
|
|
|
|
}
|
2020-06-22 23:03:55 +02:00
|
|
|
if (len >= sizeof(bs->exact_filename)) {
|
2020-06-08 20:26:38 +02:00
|
|
|
/* Name is too long to represent exactly, so leave it empty. */
|
|
|
|
bs->exact_filename[0] = '\0';
|
nbd: Fix filename generation
Export names may be used with nbd+unix, too, fix nbd_refresh_filename()
accordingly. Also, for nbd+tcp, the documented path schema is
"nbd://host[:port]/export", so use it. Furthermore, as can be seen from
that schema, the port is optional.
That makes six single cases for how the filename can be formatted; it is
not easy to generalize these cases without the resulting statement being
completely unreadable, thus there is simply one snprintf() per case.
Finally, taking the options from BDRVNBDState::socket_opts is wrong,
because those will not contain the export name. Just use
BlockDriverState::options instead.
Reported-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2014-10-08 19:55:15 +02:00
|
|
|
}
|
2014-07-18 20:24:59 +02:00
|
|
|
}
|
|
|
|
|
2019-02-01 20:29:21 +01:00
|
|
|
static char *nbd_dirname(BlockDriverState *bs, Error **errp)
|
|
|
|
{
|
|
|
|
/* The generic bdrv_dirname() implementation is able to work out some
|
|
|
|
* directory name for NBD nodes, but that would be wrong. So far there is no
|
|
|
|
* specification for how "export paths" would work, so NBD does not have
|
|
|
|
* directory names. */
|
|
|
|
error_setg(errp, "Cannot generate a base directory for NBD nodes");
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2019-02-01 20:29:25 +01:00
|
|
|
static const char *const nbd_strong_runtime_opts[] = {
|
|
|
|
"path",
|
|
|
|
"host",
|
|
|
|
"port",
|
|
|
|
"export",
|
|
|
|
"tls-creds",
|
|
|
|
"server.",
|
|
|
|
|
|
|
|
NULL
|
|
|
|
};
|
|
|
|
|
2021-02-05 17:37:12 +01:00
|
|
|
static void nbd_cancel_in_flight(BlockDriverState *bs)
|
|
|
|
{
|
|
|
|
BDRVNBDState *s = (BDRVNBDState *)bs->opaque;
|
|
|
|
|
|
|
|
reconnect_delay_timer_del(s);
|
|
|
|
|
|
|
|
if (s->state == NBD_CLIENT_CONNECTING_WAIT) {
|
|
|
|
s->state = NBD_CLIENT_CONNECTING_NOWAIT;
|
|
|
|
qemu_co_queue_restart_all(&s->free_sema);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-05-10 00:03:42 +02:00
|
|
|
static BlockDriver bdrv_nbd = {
|
2014-05-08 16:34:43 +02:00
|
|
|
.format_name = "nbd",
|
|
|
|
.protocol_name = "nbd",
|
|
|
|
.instance_size = sizeof(BDRVNBDState),
|
|
|
|
.bdrv_parse_filename = nbd_parse_filename,
|
2020-03-26 02:12:18 +01:00
|
|
|
.bdrv_co_create_opts = bdrv_co_create_opts_simple,
|
|
|
|
.create_opts = &bdrv_create_opts_simple,
|
2014-05-08 16:34:43 +02:00
|
|
|
.bdrv_file_open = nbd_open,
|
2019-09-30 23:38:20 +02:00
|
|
|
.bdrv_reopen_prepare = nbd_client_reopen_prepare,
|
2016-07-16 01:23:07 +02:00
|
|
|
.bdrv_co_preadv = nbd_client_co_preadv,
|
|
|
|
.bdrv_co_pwritev = nbd_client_co_pwritev,
|
2016-10-14 20:33:18 +02:00
|
|
|
.bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes,
|
2014-05-08 16:34:43 +02:00
|
|
|
.bdrv_close = nbd_close,
|
|
|
|
.bdrv_co_flush_to_os = nbd_co_flush,
|
2016-07-16 01:23:02 +02:00
|
|
|
.bdrv_co_pdiscard = nbd_client_co_pdiscard,
|
2015-02-06 12:24:43 +01:00
|
|
|
.bdrv_refresh_limits = nbd_refresh_limits,
|
2020-07-27 23:58:43 +02:00
|
|
|
.bdrv_co_truncate = nbd_co_truncate,
|
2014-05-08 16:34:43 +02:00
|
|
|
.bdrv_getlength = nbd_getlength,
|
2019-06-11 12:27:19 +02:00
|
|
|
.bdrv_detach_aio_context = nbd_client_detach_aio_context,
|
|
|
|
.bdrv_attach_aio_context = nbd_client_attach_aio_context,
|
2019-10-09 10:41:57 +02:00
|
|
|
.bdrv_co_drain_begin = nbd_client_co_drain_begin,
|
|
|
|
.bdrv_co_drain_end = nbd_client_co_drain_end,
|
2014-07-18 20:24:59 +02:00
|
|
|
.bdrv_refresh_filename = nbd_refresh_filename,
|
2018-03-12 16:21:23 +01:00
|
|
|
.bdrv_co_block_status = nbd_client_co_block_status,
|
2019-02-01 20:29:21 +01:00
|
|
|
.bdrv_dirname = nbd_dirname,
|
2019-02-01 20:29:25 +01:00
|
|
|
.strong_runtime_opts = nbd_strong_runtime_opts,
|
2021-02-05 17:37:12 +01:00
|
|
|
.bdrv_cancel_in_flight = nbd_cancel_in_flight,
|
2012-11-04 13:04:24 +01:00
|
|
|
};
|
|
|
|
|
|
|
|
static BlockDriver bdrv_nbd_tcp = {
|
2014-05-08 16:34:43 +02:00
|
|
|
.format_name = "nbd",
|
|
|
|
.protocol_name = "nbd+tcp",
|
|
|
|
.instance_size = sizeof(BDRVNBDState),
|
|
|
|
.bdrv_parse_filename = nbd_parse_filename,
|
2020-03-26 02:12:18 +01:00
|
|
|
.bdrv_co_create_opts = bdrv_co_create_opts_simple,
|
|
|
|
.create_opts = &bdrv_create_opts_simple,
|
2014-05-08 16:34:43 +02:00
|
|
|
.bdrv_file_open = nbd_open,
|
2019-09-30 23:38:20 +02:00
|
|
|
.bdrv_reopen_prepare = nbd_client_reopen_prepare,
|
2016-07-16 01:23:07 +02:00
|
|
|
.bdrv_co_preadv = nbd_client_co_preadv,
|
|
|
|
.bdrv_co_pwritev = nbd_client_co_pwritev,
|
2016-10-14 20:33:18 +02:00
|
|
|
.bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes,
|
2014-05-08 16:34:43 +02:00
|
|
|
.bdrv_close = nbd_close,
|
|
|
|
.bdrv_co_flush_to_os = nbd_co_flush,
|
2016-07-16 01:23:02 +02:00
|
|
|
.bdrv_co_pdiscard = nbd_client_co_pdiscard,
|
2015-02-06 12:24:43 +01:00
|
|
|
.bdrv_refresh_limits = nbd_refresh_limits,
|
2020-07-27 23:58:43 +02:00
|
|
|
.bdrv_co_truncate = nbd_co_truncate,
|
2014-05-08 16:34:43 +02:00
|
|
|
.bdrv_getlength = nbd_getlength,
|
2019-06-11 12:27:19 +02:00
|
|
|
.bdrv_detach_aio_context = nbd_client_detach_aio_context,
|
|
|
|
.bdrv_attach_aio_context = nbd_client_attach_aio_context,
|
2019-10-09 10:41:57 +02:00
|
|
|
.bdrv_co_drain_begin = nbd_client_co_drain_begin,
|
|
|
|
.bdrv_co_drain_end = nbd_client_co_drain_end,
|
2014-07-18 20:24:59 +02:00
|
|
|
.bdrv_refresh_filename = nbd_refresh_filename,
|
2018-03-12 16:21:23 +01:00
|
|
|
.bdrv_co_block_status = nbd_client_co_block_status,
|
2019-02-01 20:29:21 +01:00
|
|
|
.bdrv_dirname = nbd_dirname,
|
2019-02-01 20:29:25 +01:00
|
|
|
.strong_runtime_opts = nbd_strong_runtime_opts,
|
2021-02-05 17:37:12 +01:00
|
|
|
.bdrv_cancel_in_flight = nbd_cancel_in_flight,
|
2012-11-04 13:04:24 +01:00
|
|
|
};
|
|
|
|
|
|
|
|
static BlockDriver bdrv_nbd_unix = {
|
2014-05-08 16:34:43 +02:00
|
|
|
.format_name = "nbd",
|
|
|
|
.protocol_name = "nbd+unix",
|
|
|
|
.instance_size = sizeof(BDRVNBDState),
|
|
|
|
.bdrv_parse_filename = nbd_parse_filename,
|
2020-03-26 02:12:18 +01:00
|
|
|
.bdrv_co_create_opts = bdrv_co_create_opts_simple,
|
|
|
|
.create_opts = &bdrv_create_opts_simple,
|
2014-05-08 16:34:43 +02:00
|
|
|
.bdrv_file_open = nbd_open,
|
2019-09-30 23:38:20 +02:00
|
|
|
.bdrv_reopen_prepare = nbd_client_reopen_prepare,
|
2016-07-16 01:23:07 +02:00
|
|
|
.bdrv_co_preadv = nbd_client_co_preadv,
|
|
|
|
.bdrv_co_pwritev = nbd_client_co_pwritev,
|
2016-10-14 20:33:18 +02:00
|
|
|
.bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes,
|
2014-05-08 16:34:43 +02:00
|
|
|
.bdrv_close = nbd_close,
|
|
|
|
.bdrv_co_flush_to_os = nbd_co_flush,
|
2016-07-16 01:23:02 +02:00
|
|
|
.bdrv_co_pdiscard = nbd_client_co_pdiscard,
|
2015-02-06 12:24:43 +01:00
|
|
|
.bdrv_refresh_limits = nbd_refresh_limits,
|
2020-07-27 23:58:43 +02:00
|
|
|
.bdrv_co_truncate = nbd_co_truncate,
|
2014-05-08 16:34:43 +02:00
|
|
|
.bdrv_getlength = nbd_getlength,
|
2019-06-11 12:27:19 +02:00
|
|
|
.bdrv_detach_aio_context = nbd_client_detach_aio_context,
|
|
|
|
.bdrv_attach_aio_context = nbd_client_attach_aio_context,
|
2019-10-09 10:41:57 +02:00
|
|
|
.bdrv_co_drain_begin = nbd_client_co_drain_begin,
|
|
|
|
.bdrv_co_drain_end = nbd_client_co_drain_end,
|
2014-07-18 20:24:59 +02:00
|
|
|
.bdrv_refresh_filename = nbd_refresh_filename,
|
2018-03-12 16:21:23 +01:00
|
|
|
.bdrv_co_block_status = nbd_client_co_block_status,
|
2019-02-01 20:29:21 +01:00
|
|
|
.bdrv_dirname = nbd_dirname,
|
2019-02-01 20:29:25 +01:00
|
|
|
.strong_runtime_opts = nbd_strong_runtime_opts,
|
2021-02-05 17:37:12 +01:00
|
|
|
.bdrv_cancel_in_flight = nbd_cancel_in_flight,
|
2008-07-03 15:41:03 +02:00
|
|
|
};
|
2009-05-10 00:03:42 +02:00
|
|
|
|
|
|
|
static void bdrv_nbd_init(void)
|
|
|
|
{
|
|
|
|
bdrv_register(&bdrv_nbd);
|
2012-11-04 13:04:24 +01:00
|
|
|
bdrv_register(&bdrv_nbd_tcp);
|
|
|
|
bdrv_register(&bdrv_nbd_unix);
|
2009-05-10 00:03:42 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
block_init(bdrv_nbd_init);
|