This patch implements active synchronous mirroring. In active mode, the
passive mechanism will still be in place and is used to copy all
initially dirty clusters off the source disk; but every write request
will write data both to the source and the target disk, so the source
cannot be dirtied faster than data is mirrored to the target. Also,
once the block job has converged (BLOCK_JOB_READY sent), source and
target are guaranteed to stay in sync (unless an error occurs).
Active mode is completely optional and currently disabled at runtime. A
later patch will add a way for users to enable it.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Message-id: 20180613181823.13618-13-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-id: 20180613181823.13618-12-mreitz@redhat.com
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
This will allow us to access the block job data when the mirror block
driver becomes more complex.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Message-id: 20180613181823.13618-11-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
This new function allows to look for a consecutively dirty area in a
dirty bitmap.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20180613181823.13618-10-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
Add a function that wraps hbitmap_iter_next() and always calls it in
non-advancing mode first, and in advancing mode next. The result should
always be the same.
By using this function everywhere we called hbitmap_iter_next() before,
we should get good test coverage for non-advancing hbitmap_iter_next().
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20180613181823.13618-9-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
This new parameter allows the caller to just query the next dirty
position without moving the iterator.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 20180613181823.13618-8-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
Currently, bdrv_replace_node() refuses to create loops from one BDS to
itself if the BDS to be replaced is the backing node of the BDS to
replace it: Say there is a node A and a node B. Replacing B by A means
making all references to B point to A. If B is a child of A (i.e. A has
a reference to B), that would mean we would have to make this reference
point to A itself -- so we'd create a loop.
bdrv_replace_node() (through should_update_child()) refuses to do so if
B is the backing node of A. There is no reason why we should create
loops if B is not the backing node of A, though. The BDS graph should
never contain loops, so we should always refuse to create them.
If B is a child of A and B is to be replaced by A, we should simply
leave B in place there because it is the most sensible choice.
A more specific argument would be: Putting filter drivers into the BDS
graph is basically the same as appending an overlay to a backing chain.
But the main child BDS of a filter driver is not "backing" but "file",
so restricting the no-loop rule to backing nodes would fail here.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 20180613181823.13618-7-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
With this, the mirror_top_bs is no longer just a technically required
node in the BDS graph but actually represents the block job operation.
Also, drop MirrorBlockJob.source, as we can reach it through
mirror_top_bs->backing.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 20180613181823.13618-6-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
This patch makes the mirror code differentiate between simply waiting
for any operation to complete (mirror_wait_for_free_in_flight_slot())
and specifically waiting for all operations touching a certain range of
the virtual disk to complete (mirror_wait_on_conflicts()).
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Message-id: 20180613181823.13618-5-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
Attach a CoQueue to each in-flight operation so if we need to wait for
any we can use it to wait instead of just blindly yielding and hoping
for some operation to wake us.
A later patch will use this infrastructure to allow requests accessing
the same area of the virtual disk to specifically wait for each other.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Message-id: 20180613181823.13618-4-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
In order to talk to the source BDS (and maybe in the future to the
target BDS as well) directly, we need to convert our existing AIO
requests into coroutine I/O requests.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Message-id: 20180613181823.13618-3-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
When converting mirror's I/O to coroutines, we are going to need a point
where these coroutines are created. mirror_perform() is going to be
that point.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 20180613181823.13618-2-mreitz@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
Introduce a new global big lock for mon_fdsets. Take it where needed.
The monitor_fdset_get_fd() handling is a bit tricky: now we need to call
qemu_mutex_unlock() which might pollute errno, so we need to make sure
the correct errno be passed up to the callers. To make things simpler,
we let monitor_fdset_get_fd() return the -errno directly when error
happens, then in qemu_open() we move it back into errno.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180608035511.7439-8-peterx@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Before this patch, monitor fd helpers might be called even earlier than
monitor_init_globals(). This can be problematic.
After previous work, now monitor_init_globals() does not depend on
accelerator initialization any more. Call it earlier (before CLI
parsing; that's where the monitor APIs might be called) to make sure it
is called before any of the monitor APIs.
Suggested-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180608035511.7439-7-peterx@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Instead, use a dynamic function to detect which clock we'll use. The
problem is that the old code will let monitor initialization depend on
configure_accelerator() (that's where qtest_enabled() start to take
effect). After this change, we don't have such a dependency any more.
We just need to make sure configure_accelerator() is called when we
start to use it. Now it's only used in monitor_qapi_event_queue() and
monitor_qapi_event_handler(), so we're good.
Suggested-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180608035511.7439-6-peterx@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
[monitor_get_event_clock() name and comment tweaked]
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Fix typo in d622cb5879. Meanwhile move these variables close to each
other. monitor_qapi_event_state can be declared static, add that.
Reported-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180608035511.7439-5-peterx@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Add some explicit comments for both Readline and cpu_set/cpu_get helpers
that they do not need the mon_lock protection.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180608035511.7439-4-peterx@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
mon->fds were protected by BQL. Now protect it by mon_lock so that it
can even be used in monitor iothread.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180608035511.7439-3-peterx@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
The out_lock is protecting a few Monitor fields. In the future the
monitor code will start to run in multiple threads. We are going to
turn it into a bigger lock to protect not only the out buffer but also
most of the rest.
Since at it, rearrange the Monitor struct a bit.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20180608035511.7439-2-peterx@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
The -O2 optimization flag is passed via CFLAGS to the firmware Makefile,
but in netbook.mak, we've got some rules that only use QEMU_CFLAGS for
compiling the libc and libnet from SLOF, so these files get compiled
without optimization so far. Use CFLAGS here, too, to create faster
and smaller code.
We can additionally save some more bytes in the firmware images by compi-
ling the code with -fno-asynchronous-unwind-tables. This will omit some
ELF sections (used for stack unwinding for example) from the image that
we do not need in the firmware.
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
With the STSI instruction, we can get the UUID of the current VM instance,
so we can support loading pxelinux config files via UUID in the file name,
too.
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Tested-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Since it is quite cumbersome to manually create a combined kernel with
initrd image for network booting, we now support loading via pxelinux
configuration files, too. In these files, the kernel, initrd and command
line parameters can be specified seperately, and the firmware then takes
care of glueing everything together in memory after the files have been
downloaded. See this URL for details about the config file layout:
https://www.syslinux.org/wiki/index.php?title=PXELINUX
The user can either specify a config file directly as bootfile via DHCP
(but in this case, the file has to start either with "default" or a "#"
comment so we can distinguish it from binary kernels), or a folder (i.e.
the bootfile name must end with "/") where the firmware should look for
the typical pxelinux.cfg file names, e.g. based on MAC or IP address.
We also support the pxelinux.cfg DHCP options 209 and 210 from RFC 5071.
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Tested-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
The ip_version information now has to be stored in the filename_ip_t
structure, and there is now a common function called tftp_get_error_info()
which can be used to get the error string for a TFTP error code.
We can also get rid of some superfluous "(char *)" casts now.
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Tested-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Loadparm is defined by the s390 architecture to be 8 bytes
in length. Let's define this size in the s390-ccw bios.
Suggested-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Removing a drive with drive_del while it is being used to run an I/O
intensive workload can cause QEMU to crash.
An AIO flush can yield at some point:
blk_aio_flush_entry()
blk_co_flush(blk)
bdrv_co_flush(blk->root->bs)
...
qemu_coroutine_yield()
and let the HMP command to run, free blk->root and give control
back to the AIO flush:
hmp_drive_del()
blk_remove_bs()
bdrv_root_unref_child(blk->root)
child_bs = blk->root->bs
bdrv_detach_child(blk->root)
bdrv_replace_child(blk->root, NULL)
blk->root->bs = NULL
g_free(blk->root) <============== blk->root becomes stale
bdrv_unref(child_bs)
bdrv_delete(child_bs)
bdrv_close()
bdrv_drained_begin()
bdrv_do_drained_begin()
bdrv_drain_recurse()
aio_poll()
...
qemu_coroutine_switch()
and the AIO flush completion ends up dereferencing blk->root:
blk_aio_complete()
scsi_aio_complete()
blk_get_aio_context(blk)
bs = blk_bs(blk)
ie, bs = blk->root ? blk->root->bs : NULL
^^^^^
stale
The problem is that we should avoid making block driver graph
changes while we have in-flight requests. Let's drain all I/O
for this BB before calling bdrv_root_unref_child().
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This tests both adding and remove a node between bdrv_drain_all_begin()
and bdrv_drain_all_end(), and enabled the existing detach test for
drain_all.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
bdrv_drain_all_*() used bdrv_next() to iterate over all root nodes and
did a subtree drain for each of them. This works fine as long as the
graph is static, but sadly, reality looks different.
If the graph changes so that root nodes are added or removed, we would
have to compensate for this. bdrv_next() returns each root node only
once even if it's the root node for multiple BlockBackends or for a
monitor-owned block driver tree, which would only complicate things.
The much easier and more obviously correct way is to fundamentally
change the way the functions work: Iterate over all BlockDriverStates,
no matter who owns them, and drain them individually. Compensation is
only necessary when a new BDS is created inside a drain_all section.
Removal of a BDS doesn't require any action because it's gone afterwards
anyway.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
In the future, bdrv_drained_all_begin/end() will drain all invidiual
nodes separately rather than whole subtrees. This means that we don't
want to propagate the drain to all parents any more: If the parent is a
BDS, it will already be drained separately. Recursing to all parents is
unnecessary work and would make it an O(n²) operation.
Prepare the drain function for the changed drain_all by adding an
ignore_bds_parents parameter to the internal implementation that
prevents the propagation of the drain to BDS parents. We still (have to)
propagate it to non-BDS parents like BlockBackends or Jobs because those
are not drained separately.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Before we can introduce a single polling loop for all nodes in
bdrv_drain_all_begin(), we must make sure to run it outside of coroutine
context like we already do for bdrv_do_drained_begin().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
bdrv_drain_all() wants to have a single polling loop for draining the
in-flight requests of all nodes. This means that the AIO_WAIT_WHILE()
condition relies on activity in multiple AioContexts, which is polled
from the mainloop context. We must therefore call AIO_WAIT_WHILE() from
the mainloop thread and use the AioWait notification mechanism.
Just randomly picking the AioContext of any non-mainloop thread would
work, but instead of bothering to find such a context in the caller, we
can just as well accept NULL for ctx.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We cannot allow aio_poll() in bdrv_drain_invoke(begin=true) until we're
done with propagating the drain through the graph and are doing the
single final BDRV_POLL_WHILE().
Just schedule the coroutine with the callback and increase bs->in_flight
to make sure that the polling phase will wait for it.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
bdrv_do_drained_begin() is only safe if we have a single
BDRV_POLL_WHILE() after quiescing all affected nodes. We cannot allow
that parent callbacks introduce a nested polling loop that could cause
graph changes while we're traversing the graph.
Split off bdrv_do_drained_begin_quiesce(), which only quiesces a single
node without waiting for its requests to complete. These requests will
be waited for in the BDRV_POLL_WHILE() call down the call chain.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
If bdrv_do_drained_begin() polls during its subtree recursion, the graph
can change and mess up the bs->children iteration. Test that this
doesn't happen.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Anything can happen inside BDRV_POLL_WHILE(), including graph
changes that may interfere with its callers (e.g. child list iteration
in recursive callers of bdrv_do_drained_begin).
Switch to a single BDRV_POLL_WHILE() call for the whole subtree at the
end of bdrv_do_drained_begin() to avoid such effects. The recursion
happens now inside the loop condition. As the graph can only change
between bdrv_drain_poll() calls, but not inside of it, doing the
recursion here is safe.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch adds two bdrv-drain tests for what happens if some BDS goes
away during the drainage.
The basic idea is that you have a parent BDS with some child nodes.
Then, you drain one of the children. Because of that, the party who
actually owns the parent decides to (A) delete it, or (B) detach all its
children from it -- both while the child is still being drained.
A real-world case where this can happen is the mirror block job, which
may exit if you drain one of its children.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
For bdrv_drain(), recursively waiting for child node requests is
pointless because we didn't quiesce their parents, so new requests could
come in anyway. Letting the function work only on a single node makes it
more consistent.
For subtree drains and drain_all, we already have the recursion in
bdrv_do_drained_begin(), so the extra recursion doesn't add anything
either.
Remove the useless code.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
We already requested that block jobs be paused in .bdrv_drained_begin,
but no guarantee was made that the job was actually inactive at the
point where bdrv_drained_begin() returned.
This introduces a new callback BdrvChildRole.bdrv_drained_poll() and
uses it to make bdrv_drain_poll() consider block jobs using the node to
be drained.
For the test case to work as expected, we have to switch from
block_job_sleep_ns() to qemu_co_sleep_ns() so that the test job is even
considered active and must be waited for when draining the node.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Commit 91af091f92 added an additional aio_poll() to BDRV_POLL_WHILE()
in order to make sure that all pending BHs are executed on drain. This
was the wrong place to make the fix, as it is useless overhead for all
other users of the macro and unnecessarily complicates the mechanism.
This patch effectively reverts said commit (the context has changed a
bit and the code has moved to AIO_WAIT_WHILE()) and instead polls in the
loop condition for drain.
The effect is probably hard to measure in any real-world use case
because actual I/O will dominate, but if I run only the initialisation
part of 'qemu-img convert' where it calls bdrv_block_status() for the
whole image to find out how much data there is copy, this phase actually
needs only roughly half the time after this patch.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Since we use bdrv_do_drained_begin/end() for bdrv_drain_all_begin/end(),
coroutine context is automatically left with a BH, preventing the
deadlocks that made bdrv_drain_all*() unsafe in coroutine context. Now
that we even removed the old polling code as dead code, it's obvious
that it's compatible now.
Enable the coroutine test cases for bdrv_drain_all().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
All involved nodes are already idle, we called bdrv_do_drain_begin() on
them.
The comment in the code suggested that this was not correct because the
completion of a request on one node could spawn a new request on a
different node (which might have been drained before, so we wouldn't
drain the new request). In reality, new requests to different nodes
aren't spawned out of nothing, but only in the context of a parent
request, and they aren't submitted to random nodes, but only to child
nodes. As long as we still poll for the completion of the parent request
(which we do), draining each root node separately is good enough.
Remove the additional polling code from bdrv_drain_all_begin() and
replace it with an assertion that all nodes are already idle after we
drained them separately.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
All callers pass false for the 'recursive' parameter now. Remove it.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
bdrv_do_drain_begin/end() implement already everything that
bdrv_drain_all_begin/end() need and currently still do manually: Disable
external events, call parent drain callbacks, call block driver
callbacks.
It also does two more things:
The first is incrementing bs->quiesce_counter. bdrv_drain_all() already
stood out in the test case by behaving different from the other drain
variants. Adding this is not only safe, but in fact a bug fix.
The second is calling bdrv_drain_recurse(). We already do that later in
the same function in a loop, so basically doing an early first iteration
doesn't hurt.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
As long as nobody keeps the other I/O thread from working, there is no
reason why bdrv_drain() wouldn't work with cross-AioContext events. The
key is that the root request we're waiting for is in the AioContext
we're polling (which it always is for bdrv_drain()) so that aio_poll()
is woken up in the end.
Add a test case that shows that it works. Remove the comment in
bdrv_drain() that claims otherwise.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
In commit 802cbcb730, most issues have been fixed when qemu guest
migration. But the queue size still need to check whether is equal to
PS2_QUEUE_SIZE. If yes, the wptr should set as 0. Or, wptr would larger
than PS2_QUEUE_SIZE and never come back when ps2_queue_noirq is called.
This could lead to OOB access, add check to avoid it.
Signed-off-by: liujunjie <liujunjie23@huawei.com>
Reviewed-by: Gonglei <arei.gonglei@huawei.com>
Message-id: 20180607080237.12360-1-liujunjie23@huawei.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
