Add a new vm command MIG_CMD_RECV_BITMAP to request received bitmap for
one ramblock.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-12-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
On the destination side, we cannot wake up all the threads when we got
reconnected. The first thing to do is to wake up the main load thread,
so that we can continue to receive valid messages from source again and
reply when needed.
At this point, we switch the destination VM state from postcopy-paused
back to postcopy-recover.
Now we are finally ready to do the resume logic.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-11-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Introducing new migration state "postcopy-recover". If a migration
procedure is paused and the connection is rebuilt afterward
successfully, we'll switch the source VM state from "postcopy-paused" to
the new state "postcopy-recover", then we'll do the resume logic in the
migration thread (along with the return path thread).
This patch only do the state switch on source side. Another following up
patch will handle the state switching on destination side using the same
status bit.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-10-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
s/2.11/2.13/
This patch detects the "resume" flag of migration command, rebuild the
channels only if the flag is set.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-9-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
It will be used when we want to resume one paused migration.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-8-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
---
s/2.12/2.13/
Allows the fault thread to stop handling page faults temporarily. When
network failure happened (and if we expect a recovery afterwards), we
should not allow the fault thread to continue sending things to source,
instead, it should halt for a while until the connection is rebuilt.
When the dest main thread noticed the failure, it kicks the fault thread
to switch to pause state.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-7-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Let the thread pause for network issues.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-6-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
When there is IO error on the incoming channel (e.g., network down),
instead of bailing out immediately, we allow the dst vm to switch to the
new POSTCOPY_PAUSE state. Currently it is still simple - it waits the
new semaphore, until someone poke it for another attempt.
One note is that here on ram loading thread we cannot detect the
POSTCOPY_ACTIVE state, but we need to detect the more specific
POSTCOPY_INCOMING_RUNNING state, to make sure we have already loaded all
the device states.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-5-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Now when network down for postcopy, the source side will not fail the
migration. Instead we convert the status into this new paused state, and
we will try to wait for a rescue in the future.
If a recovery is detected, migration_thread() will reset its local
variables to prepare for that.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-4-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Introducing a new state "postcopy-paused", which can be used when the
postcopy migration is paused. It is targeted for postcopy network
failure recovery.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-3-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The old incoming migration is running in main thread and default
gcontext. With the new qio_channel_add_watch_full() we can now let it
run in the thread's own gcontext (if there is one).
Currently this patch does nothing alone. But when any of the incoming
migration is run in another iothread (e.g., the upcoming migrate-recover
command), this patch will bind the incoming logic to the iothread
instead of the main thread (which may already get page faulted and
hanged).
RDMA is not considered for now since it's not even using the QIO watch
framework at all.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180502104740.12123-2-peterx@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Once there, we don't need the struct names anywhere, just the
typedefs. And now also document all fields.
Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
--
Be network agnostic.
Add error checking for all values.
We need to make sure that we have started all the multifd threads.
Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
In both sides. We still don't transmit anything through them.
Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Once there, make count field to always be accessed with atomic
operations. To make blocking operations, we need to know that the
thread is running, so create a bool to indicate that.
Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
--
Once here, s/terminate_multifd_*-threads/multifd_*_terminate_threads/
This is consistente with every other function
Fix the bug introduced by da3f56cb2e (migration: remove
ram_save_compressed_page()), It should be 'return' rather than
'res'
Sorry for this stupid mistake :(
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180428081045.8878-1-xiaoguangrong@tencent.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Release buf on error path too.
Bug was introduced in b35ebdf076 "migration: add postcopy
migration of dirty bitmaps" with the whole function.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20180427142002.21930-3-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
CC: qemu-stable@nongnu.org
Signed-off-by: Eric Blake <eblake@redhat.com>
Now that we can safely call QOBJECT() on QObject * as well as its
subtypes, we can have macros qobject_ref() / qobject_unref() that work
everywhere instead of having to use QINCREF() / QDECREF() for QObject
and qobject_incref() / qobject_decref() for its subtypes.
The replacement is mechanical, except I broke a long line, and added a
cast in monitor_qmp_cleanup_req_queue_locked(). Unlike
qobject_decref(), qobject_unref() doesn't accept void *.
Note that the new macros evaluate their argument exactly once, thus no
need to shout them.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20180419150145.24795-4-marcandre.lureau@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[Rebased, semantic conflict resolved, commit message improved]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Now, we can reuse the path in ram_save_page() to post the page out
as normal, then the only thing remained in ram_save_compressed_page()
is compression that we can move it out to the caller
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180330075128.26919-11-xiaoguangrong@tencent.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
It directly sends the page to the stream neither checking zero nor
using xbzrle or compression
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180330075128.26919-10-xiaoguangrong@tencent.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
save_zero_page() is always our first approach to try, move it to
the common place before calling ram_save_compressed_page
and ram_save_page
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180330075128.26919-9-xiaoguangrong@tencent.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
The function is called by both ram_save_page and ram_save_target_page,
so move it to the common caller to cleanup the code
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180330075128.26919-8-xiaoguangrong@tencent.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Move some code from ram_save_target_page() to ram_save_host_page()
to make it be more readable for latter patches that dramatically
clean ram_save_target_page() up
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180330075128.26919-7-xiaoguangrong@tencent.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Abstract the common function control_save_page() to cleanup the code,
no logic is changed
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180330075128.26919-6-xiaoguangrong@tencent.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Currently the page being compressed is allowed to be updated by
the VM on the source QEMU, correspondingly the destination QEMU
just ignores the decompression error. However, we completely miss
the chance to catch real errors, then the VM is corrupted silently
To make the migration more robuster, we copy the page to a buffer
first to avoid it being written by VM, then detect and handle the
errors of both compression and decompression errors properly
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180330075128.26919-5-xiaoguangrong@tencent.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Current code uses uncompress() to decompress memory which manages
memory internally, that causes huge memory is allocated and freed
very frequently, more worse, frequently returning memory to kernel
will flush TLBs
So, we maintain the memory by ourselves and reuse it for each
decompression
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Jiang Biao <jiang.biao2@zte.com.cn>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180330075128.26919-4-xiaoguangrong@tencent.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Current code uses compress2() to compress memory which manages memory
internally, that causes huge memory is allocated and freed very
frequently
More worse, frequently returning memory to kernel will flush TLBs
and trigger invalidation callbacks on mmu-notification which
interacts with KVM MMU, that dramatically reduce the performance
of VM
So, we maintain the memory by ourselves and reuse it for each
compression
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Jiang Biao <jiang.biao2@zte.com.cn>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180330075128.26919-3-xiaoguangrong@tencent.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
As compression is a heavy work, do not do it in migration thread,
instead, we post it out as a normal page
Reviewed-by: Wei Wang <wei.w.wang@intel.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Xiao Guangrong <xiaoguangrong@tencent.com>
Message-Id: <20180330075128.26919-2-xiaoguangrong@tencent.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Postcopy total blocktime is available on destination side only.
But query-migrate was possible only for source. This patch
adds ability to call query-migrate on destination.
To be able to see postcopy blocktime, need to request postcopy-blocktime
capability.
The query-migrate command will show following sample result:
{"return":
"postcopy-vcpu-blocktime": [115, 100],
"status": "completed",
"postcopy-blocktime": 100
}}
postcopy_vcpu_blocktime contains list, where the first item is the first
vCPU in QEMU.
This patch has a drawback, it combines states of incoming and
outgoing migration. Ongoing migration state will overwrite incoming
state. Looks like better to separate query-migrate for incoming and
outgoing migration or add parameter to indicate type of migration.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Alexey Perevalov <a.perevalov@samsung.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Message-Id: <1521742647-25550-7-git-send-email-a.perevalov@samsung.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
This patch provides blocktime calculation per vCPU,
as a summary and as a overlapped value for all vCPUs.
This approach was suggested by Peter Xu, as an improvements of
previous approch where QEMU kept tree with faulted page address and cpus bitmask
in it. Now QEMU is keeping array with faulted page address as value and vCPU
as index. It helps to find proper vCPU at UFFD_COPY time. Also it keeps
list for blocktime per vCPU (could be traced with page_fault_addr)
Blocktime will not calculated if postcopy_blocktime field of
MigrationIncomingState wasn't initialized.
Signed-off-by: Alexey Perevalov <a.perevalov@samsung.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Message-Id: <1521742647-25550-4-git-send-email-a.perevalov@samsung.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
This patch adds request to kernel space for UFFD_FEATURE_THREAD_ID, in
case this feature is provided by kernel.
PostcopyBlocktimeContext is encapsulated inside postcopy-ram.c,
due to it being a postcopy-only feature.
Also it defines PostcopyBlocktimeContext's instance live time.
Information from PostcopyBlocktimeContext instance will be provided
much after postcopy migration end, instance of PostcopyBlocktimeContext
will live till QEMU exit, but part of it (vcpu_addr,
page_fault_vcpu_time) used only during calculation, will be released
when postcopy ended or failed.
To enable postcopy blocktime calculation on destination, need to
request proper compatibility (Patch for documentation will be at the
tail of the patch set).
As an example following command enable that capability, assume QEMU was
started with
-chardev socket,id=charmonitor,path=/var/lib/migrate-vm-monitor.sock
option to control it
[root@host]#printf "{\"execute\" : \"qmp_capabilities\"}\r\n \
{\"execute\": \"migrate-set-capabilities\" , \"arguments\": {
\"capabilities\": [ { \"capability\": \"postcopy-blocktime\", \"state\":
true } ] } }" | nc -U /var/lib/migrate-vm-monitor.sock
Or just with HMP
(qemu) migrate_set_capability postcopy-blocktime on
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Alexey Perevalov <a.perevalov@samsung.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Message-Id: <1521742647-25550-3-git-send-email-a.perevalov@samsung.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Right now it could be used on destination side to
enable vCPU blocktime calculation for postcopy live migration.
vCPU blocktime - it's time since vCPU thread was put into
interruptible sleep, till memory page was copied and thread awake.
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Alexey Perevalov <a.perevalov@samsung.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Message-Id: <1521742647-25550-2-git-send-email-a.perevalov@samsung.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
This reverts commit 0746a92612.
Discussion with kwolf suggests this is actually an API change that
we need to gate on a capability. Push to 2.13.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Activating the block devices causes the locks to be taken on
the backing file. If we're running with -S and the destination libvirt
hasn't started the destination with 'cont', it's expecting the locks are
still untaken.
Don't activate the block devices if we're not going to autostart the VM;
'cont' already will do that anyway.
bz: https://bugzilla.redhat.com/show_bug.cgi?id=1560854
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-Id: <20180328170207.49512-1-dgilbert@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Fix leak spotted by ASAN:
Direct leak of 16 byte(s) in 1 object(s) allocated from:
#0 0x7fe1abb80a38 in __interceptor_calloc (/lib64/libasan.so.4+0xdea38)
#1 0x7fe1aaf1bf75 in g_malloc0 ../glib/gmem.c:124
#2 0x7fe1aaf1c249 in g_malloc0_n ../glib/gmem.c:355
#3 0x55f4841cfaa9 in postcopy_ram_fault_thread /home/elmarco/src/qemu/migration/postcopy-ram.c:596
#4 0x55f48479447b in qemu_thread_start /home/elmarco/src/qemu/util/qemu-thread-posix.c:504
#5 0x7fe1a043550a in start_thread (/lib64/libpthread.so.0+0x750a)
Regression introduced with commit 00fa4fc85b.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20180321113644.21899-1-marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Fix the case where when a migration with a bad protocol is tried,
we leave the block migration capability set.
(This is a cut down version of my 'migration: Fix block failure cases'
where it's other case was fixed by Peter's dd0ee30cae )
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-Id: <20180316202114.32345-1-dgilbert@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
only read_done blocks are in the queued to be flushed to the migration
stream. submitted blocks are still in flight.
Signed-off-by: Peter Lieven <pl@kamp.de>
Message-Id: <1520507908-16743-6-git-send-email-pl@kamp.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
the current implementation submits up to 512 I/O requests in parallel
which is much to high especially for a background task.
This patch adds a maximum limit of 16 I/O requests that can
be submitted in parallel to avoid monopolizing the I/O device.
Signed-off-by: Peter Lieven <pl@kamp.de>
Message-Id: <1520507908-16743-5-git-send-email-pl@kamp.de>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
RDMA migration implement save_page function for QEMUFile, but
ram_control_save_page do not increase bytes_xfer. So when doing
RDMA migration, it will use whole bandwidth.
Signed-off-by: Lidong Chen <lidongchen@tencent.com>
Message-Id: <1520692378-1835-1-git-send-email-lidongchen@tencent.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Instead of creating a QIOChannelSocket directly for the migration
server socket, use a QIONetListener. This provides the ability
to listen on multiple sockets at the same time, so enables
full support for IPv4/IPv6 dual stack.
For example, '$QEMU -incoming tcp::9000' now correctly listens
on both 0.0.0.0 and :: at the same time, instead of only on 0.0.0.0.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-Id: <20180312141714.7223-1-berrange@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Now that we have the mechanisms in here, allow shared memory in a
postcopy.
Note that QEMU can't tell who all the users of shared regions are
and thus can't tell whether all the users of the shared regions
have appropriate support for postcopy. Those devices that explicitly
support shared memory (e.g. vhost-user) must check, but it doesn't
stop weirder configurations causing problems.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Wire up a call to VHOST_USER_POSTCOPY_END message to the vhost clients
right before we ask the listener thread to shutdown.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Cause the vhost-user client to be woken up whenever:
a) We place a page in postcopy mode
b) We get a fault and the page has already been received
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
