On Fri, Jul 4, 2014 at 2:31 PM, Paolo Bonzini wrote:
> Il 04/07/2014 07:10, Ming Lei ha scritto:
>
>> With both the two coroutine patches and the block plug&unplug patches,
>> performance of qemu.git/master virtio-blk dataplane can recover to level
>> of
>> QEMU 2.0.
>
>
> So the excessive writes
Il 03/07/2014 20:39, David Drysdale ha scritto:
> On Thu, Jul 03, 2014 at 11:12:33AM +0200, Paolo Bonzini wrote:
>> Given Linux's previous experience with BPF filters, what do you
>> think about attaching specific BPF programs to file descriptors?
>> Then whenever a syscall is run that affects a f
Il 04/07/2014 09:02, Ming Lei ha scritto:
> So the excessive writes (to eventfd and irqfd) are not a problem? That'd be
> a relief. :)
I mean it is in a level, but your aio_notify() patch still can improve virtioblk
dataplane performance some, in my test, with 5~10K IOPS improvement,
which shoul
On Fri, Jul 4, 2014 at 2:51 PM, Paolo Bonzini wrote:
> Il 03/07/2014 20:17, Stefan Hajnoczi ha scritto:
>
>> On Thu, Jul 3, 2014 at 6:59 PM, Paolo Bonzini wrote:
>>
>> Looks like a very interesting optimization for the block layer. If we
>> can get some performance results - especially how it im
Il 04/07/2014 09:23, Ming Lei ha scritto:
I think it is good and better to go to 2.1, and it should save lots of
write syscall.
Also should regression be caused, the per thread trick may be
resorted to, which should be simple.
If we have the "right" solution (which we do, unlike the plug/unplu
On Thu, Jul 03, 2014 at 09:30:05PM -0700, Richard Henderson wrote:
> > Another one is probably not worth bothering - PERR, CTPOP, CTLZ, UNPKBx and
> > PKxB
> > don't accept literal argument. For one thing, as(1) won't let you generate
> > those, so it would have to be explicit
> > .long 0x70
On 01/07/2014 18:13, fred.kon...@greensocs.com wrote:
From: KONRAD Frederic
This fixes a bug where QEMU stall in icount mode.
It happens when a simple timer callback is created on VIRTUAL CLOCK modding
itself regularly.
The actual warping mechanism is called once and then the time didn't grow
On Thu, Jul 03, 2014 at 10:05:56PM -0600, Eric Blake wrote:
> On 07/03/2014 08:56 PM, Hu Tao wrote:
> > On Thu, Jul 03, 2014 at 06:33:28AM -0600, Eric Blake wrote:
> >> On 07/03/2014 12:10 AM, Hu Tao wrote:
> >>> This patch fixes two problems of memory-backend-file:
> >>>
> >>> 1. If user adds a me
Il 04/07/2014 09:43, Hu Tao ha scritto:
The memory size is rounded up to hpagesize when allocating memory, we
can waste at most hpagesize-1 bytes memory. I don't think it's a
problem.
In the other side, do you think we should give an error in the case like
-object memory-backend-ram,size=M(g
On Fri, Jul 04, 2014 at 01:10:39PM +0800, Ming Lei wrote:
> On Thu, Jul 3, 2014 at 9:51 PM, Stefan Hajnoczi wrote:
> > The coroutine pool reuses exited coroutines to make qemu_coroutine_create()
> > cheap. The size of the pool is capped to prevent it from hogging memory
> > after
> > a period of
Il 04/07/2014 09:30, Frederic Konrad ha scritto:
+/*
+ * In icount mode, sometimes the VCPU is blocked and an event is
needed to
+ * continue.
+ * Just warp to make the time grows and have a chance to run the
CPU.
+ */
+qemu_clock_warp(QEMU_CLOCK_VIRTUAL);
return r
On Wed, Jul 02, 2014 at 11:17:40PM +0800, yue wrote:
> i now look at qemu 2.0, i do not find rbd-api related to clone.
> if qemu support this function? and from which version?
> clone api of rbd is very simple(one api), why qemu does not implement?what is
> the reason?
rbd_snap_create()? QEMU su
* Dong, Eddie (eddie.d...@intel.com) wrote:
> > >
> > > I didn't quite understand a couple of things though, perhaps you can
> > > explain:
> > >1) If we ignore the TCP sequence number problem, in an SMP machine
> > > don't we get other randomnesses - e.g. which core completes something
> > > f
On Fri, Jul 04, 2014 at 11:06:41AM +0800, Ming Lei wrote:
> +static int ioq_submit(struct qemu_laio_state *s)
> +{
> +int ret, i = 0;
> +int len = s->io_q.idx;
> +
> +do {
> +ret = io_submit(s->ctx, len, s->io_q.iocbs);
> +} while (i++ < 3 && ret == -EAGAIN);
> +
> +/* e
When a BlockDriverState is associated with a storage controller
DeviceState we expect guest I/O. Use this opportunity to bump the
coroutine pool size by 64.
This patch ensures that the coroutine pool size scales with the number
of drives attached to the guest. It should increase coroutine pool
u
Allow coroutine users to adjust the pool size. For example, if the
guest has multiple emulated disk drives we should keep around more
coroutines.
Signed-off-by: Stefan Hajnoczi
---
include/block/coroutine.h | 11 +++
qemu-coroutine.c | 26 +++---
2 files cha
v2:
* Assert that callers never reduce pool below default size [eblake]
The coroutine pool reuses exited coroutines to make qemu_coroutine_create()
cheap. The size of the pool is capped to prevent it from hogging memory after
a period of high coroutine activity. Previously the max size was hard
Hi Stefan,
Sorry for missing your comments.
On Thu, Jul 3, 2014 at 8:24 PM, Stefan Hajnoczi wrote:
> On Wed, Jul 02, 2014 at 08:18:47PM +0800, Ming Lei wrote:
>> @@ -36,9 +38,19 @@ struct qemu_laiocb {
>> QLIST_ENTRY(qemu_laiocb) node;
>> };
>>
>> +struct laio_queue {
>
> This file doesn't
On Fri, Jul 4, 2014 at 5:18 PM, Ming Lei wrote:
>>
>> This leaks requests when ret < len. I think the loop below should be
>> used in that case to fail unsubmitted requests with -EIO.
>
> I thought about the problem before, but looks it may not return 'ret'
> which is less than
> len, follows the
Use of glue is heavily uses in audio code. I completely redesigned it
for Virtualbox and removed all the hard to understand glue code :) .
Not sure if this glue magic is such heavily used anywhere else also.
Moreover audio code uses one big monolythic big audio file audio.c .
So bringing modulari
After memory region QOMification QEMU is stricter in detecting
wrong usage of the memory region API. Here it detected a
memory_region_destroy done before the corresponding
memory_region_del_subregion; the memory_region_destroy is
done by msix_uninit_exclusive_bar, the memory_region_del_subregion
i
Stefan Hajnoczi writes:
> When a BlockDriverState is associated with a storage controller
> DeviceState we expect guest I/O. Use this opportunity to bump the
> coroutine pool size by 64.
>
> This patch ensures that the coroutine pool size scales with the number
> of drives attached to the guest.
Hi,
The commit 580b6b2aa2(dataplane: use the QEMU block layer for I/O)
introduces ~40% throughput regression on virtio-blk dataplane, and
one of causes is that submitting I/O as a batch is removed.
This patchset trys to introduce this mechanism on block, at least,
linux-aio can benefit from that.
This patch introduces three APIs so that following
patches can support queuing I/O requests and submitting them
as a batch for improving I/O performance.
Reviewed-by: Paolo Bonzini
Signed-off-by: Ming Lei
---
block.c | 31 +++
include/block/block.
Before commit 580b6b2aa2(dataplane: use the QEMU block
layer for I/O), dataplane for virtio-blk submits block
I/O as a batch.
This commit 580b6b2aa2 replaces the custom linux AIO
implementation(including submit I/O as a batch) with QEMU
block layer, but this commit causes ~40% throughput regressio
This patch implements .bdrv_io_plug, .bdrv_io_unplug and
.bdrv_flush_io_queue callbacks for linux-aio Block Drivers,
so that submitting I/O as a batch can be supported on linux-aio.
Signed-off-by: Ming Lei
---
block/linux-aio.c | 96 +++--
block/
Am 04.07.2014 um 09:27 hat Paolo Bonzini geschrieben:
> Il 04/07/2014 09:23, Ming Lei ha scritto:
> >I think it is good and better to go to 2.1, and it should save lots of
> >write syscall.
> >
> >Also should regression be caused, the per thread trick may be
> >resorted to, which should be simple.
On 04/07/2014 09:57, Paolo Bonzini wrote:
Il 04/07/2014 09:30, Frederic Konrad ha scritto:
+/*
+ * In icount mode, sometimes the VCPU is blocked and an event is
needed to
+ * continue.
+ * Just warp to make the time grows and have a chance to run the
CPU.
+ */
+qemu_clo
Stefan Hajnoczi writes:
> When a BlockDriverState is associated with a storage controller
> DeviceState we expect guest I/O. Use this opportunity to bump the
> coroutine pool size by 64.
> This patch ensures that the coroutine pool size scales with the number
> of drives attached to the guest.
> Right, but when I put a timer eg on QEMU_VIRTUAL_CLOCK the guest is stuck.
> icount_warp_rt is not called neither qemu_clock_warp(..)..
It should be. timer_mod_ns -> timerlist_rearm -> qemu_clock_warp.
> So yes as you said seems a qemu_clock_warp is missing somewhere.
>
> Shouldn't icount_wa
> We have a rather long freeze phase this time (which I think is a good
> thing). This patch fixes a regression, even if it may not be the most
> important one because it is in experimental code. But I still think that
> this time in the hard freeze is the right time to commit patches like
> this.
Am 01.07.2014 14:12, schrieb Dr. David Alan Gilbert:
> Are there any standard APIs for the heartbeet thing we can already
> tie into?
Maybe the http://www.linux-ha.org/wiki/Heartbeat daemon?
Andreas
--
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guil
[CC+=linux-api@]
Hi Andrea,
On Wed, Jul 2, 2014 at 6:50 PM, Andrea Arcangeli wrote:
> This new syscall will move anon pages across vmas, atomically and
> without touching the vmas.
Please CC linux-api on patches that change the API/ABI. (See
https://www.kernel.org/doc/man-pages/linux-api-ml.htm
The purpose of the test runner is to prepare test environment (e.g. create a
work directory, a test image, etc), execute the program under test with
parameters, indicate a test failure if the program was killed during test
execution and collect core dumps, logs and other test artifacts.
The test r
'Overall fuzzer requirements' chapter contains the current product vision and
features done and to be done. This chapter is still in progress.
v1 -> v2:
* added description for fuzzer configuration files
* added the example fuzzer.conf file
* various fixes based on the Eric's review
Signed-off
Fuzz submodule of qcow2 image generator contains fuzzing functions for image
fields.
Each fuzzing function contains list of constraints and call of a helper
function that randomly selects a fuzzed value satisfied to one of constraints.
For now constraints are only known as invalid or potentially da
__init__.py provides the public API required by the test runner
Signed-off-by: Maria Kustova
---
tests/image-fuzzer/qcow2/__init__.py | 1 +
1 file changed, 1 insertion(+)
create mode 100644 tests/image-fuzzer/qcow2/__init__.py
diff --git a/tests/image-fuzzer/qcow2/__init__.py
b/tests/image-f
This patch series introduces the image fuzzer, a tool for stability and
reliability testing.
Its approach is to run large amount of tests in background. During every test a
program (e.g. qemu-img) is called to read or modify an invalid test image.
A test image has valid inner structure defined by i
Signed-off-by: Maria Kustova
---
tests/image-fuzzer/LICENSE | 339 +
1 file changed, 339 insertions(+)
create mode 100644 tests/image-fuzzer/LICENSE
diff --git a/tests/image-fuzzer/LICENSE b/tests/image-fuzzer/LICENSE
new file mode 100644
index 00
Layout submodule of qcow2 package creates a random valid image, randomly
selects some amount of its fields, fuzzes them and write the fuzzed image to
the file.
Now only header and header extensions are generated, a remaining file is filled
by zeroes.
v1 -> v2:
* Added support of fuzzer configurat
2014-07-04 1:00 GMT+04:00 Richard W.M. Jones :
> Well I got bored with the TCG test after about 4 hours. It appears to
> hang launching the L3 guest, reproducible on two different hosts both
> running qemu 2.0.0.
In my tests qemu tcg sometimes fail when run first vm. (kernel 3.12)
--
Vasiliy T
* Dong, Eddie (eddie.d...@intel.com) wrote:
> > >
> > > Let me clarify on this issue. COLO didn't ignore the TCP sequence
> > > number, but uses a new implementation to make the sequence number to
> > > be best effort identical between the primary VM (PVM) and secondary VM
> > > (SVM). Likely, VMM
Now requests are submitted as a batch, so it is natural
to notify guest as a batch too.
This may supress interrupt notification to VM:
- in my test, decreased by ~13K/sec
Signed-off-by: Ming Lei
---
hw/block/dataplane/virtio-blk.c | 13 -
hw/block/virtio-blk.c
Hi,
The first one fixes one problem introduced recently.
The second one supresses notifications to guest.
Thanks,
--
Ming Lei
The callback has to be saved and reset in virtio_blk_data_plane_start(),
otherwise dataplane's requests will be completed in qemu aio context.
Signed-off-by: Ming Lei
---
hw/block/dataplane/virtio-blk.c |7 ---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/hw/block/datapl
On Fri, Jul 04, 2014 at 04:17:11PM +0400, Vasiliy Tolstov wrote:
> 2014-07-04 1:00 GMT+04:00 Richard W.M. Jones :
> > Well I got bored with the TCG test after about 4 hours. It appears to
> > hang launching the L3 guest, reproducible on two different hosts both
> > running qemu 2.0.0.
>
>
> In m
Forward chr_add_watch call from mux chardev to underlying
implementation.
This should fix bug #1335444
Signed-off-by: Kirill Batuzov
---
qemu-char.c |9 +
1 file changed, 9 insertions(+)
diff --git a/qemu-char.c b/qemu-char.c
index 51917de..e1bd6f5 100644
--- a/qemu-char.c
+++ b/qe
Il 04/07/2014 14:27, Ming Lei ha scritto:
Now requests are submitted as a batch, so it is natural
to notify guest as a batch too.
This may supress interrupt notification to VM:
- in my test, decreased by ~13K/sec
I don't think this can work. Requests are not completed FIFO.
What yo
Series contains a few bug fixes; you might want to consider it for 2.1.
Markus Armbruster writes:
> Issues addressed in this series:
>
> * BlockDriver method bdrv_getlength() generally returns -errno, but
> some implementations return -1 instead. Fix them [PATCH 1].
>
> * Frequent conversions
Il 04/07/2014 14:43, Kirill Batuzov ha scritto:
Forward chr_add_watch call from mux chardev to underlying
implementation.
This should fix bug #1335444
Signed-off-by: Kirill Batuzov
---
qemu-char.c |9 +
1 file changed, 9 insertions(+)
diff --git a/qemu-char.c b/qemu-char.c
index
> >
> > I didn't quite understand a couple of things though, perhaps you can
> > explain:
> >1) If we ignore the TCP sequence number problem, in an SMP machine
> > don't we get other randomnesses - e.g. which core completes something
> > first, or who wins a lock contention, so the output strea
> >
> > Let me clarify on this issue. COLO didn't ignore the TCP sequence
> > number, but uses a new implementation to make the sequence number to
> > be best effort identical between the primary VM (PVM) and secondary VM
> > (SVM). Likely, VMM has to synchronize the emulation of randomization
> >
Signed-off-by: Markus Armbruster
Reviewed-by: Fam Zheng
---
hw/block/virtio-blk.c | 24 ++--
1 file changed, 14 insertions(+), 10 deletions(-)
diff --git a/hw/block/virtio-blk.c b/hw/block/virtio-blk.c
index aec3146..d946fa9 100644
--- a/hw/block/virtio-blk.c
+++ b/hw/block/
When a device model's I/O operation fails, we execute the error
action. This lets layers above QEMU implement thin provisioning, or
attempt to correct errors before they reach the guest. But when the
I/O operation fails because it's invalid, reporting the error to the
guest is the only sensible a
When a device model's I/O operation fails, we execute the error
action. This lets layers above QEMU implement thin provisioning, or
attempt to correct errors before they reach the guest. But when the
I/O operation fails because its invalid, reporting the error to the
guest is the only sensible ac
The block layer fails such reads and writes just fine. However, they
then get treated like valid operations that fail: the error action
gets executed. Unwanted; reporting the error to the guest is the only
sensible action.
Reject them before passing them to the block layer. This bypasses the
er
The block layer fails such reads and writes just fine. However, they
then get treated like valid operations that fail: the error action
gets executed. Unwanted; reporting the error to the guest is the only
sensible action.
Reject them before passing them to the block layer. This bypasses the
er
qemu_clock_enable says:
/* Disabling the clock will wait for related timerlists to stop
* executing qemu_run_timers. Thus, this functions should not
* be used from the callback of a timer that is based on @clock.
* Doing so would cause a deadlock.
*/
and it indeed does: vm_stop uses qemu_clo
On Fri, 4 Jul 2014 15:48:50 +0200
Paolo Bonzini wrote:
> qemu_clock_enable says:
>
> /* Disabling the clock will wait for related timerlists to stop
> * executing qemu_run_timers. Thus, this functions should not
> * be used from the callback of a timer that is based on @clock.
> * Doing so
On 30 June 2014 13:09, Paolo Bonzini wrote:
> Il 28/06/2014 18:59, Peter Maydell ha scritto:
>
>> Having a direct dependency
>>Makefile: $(GENERATED_HEADERS)
>> can result in not-from-clean builds failing sometimes, because it means
>> that when Make does its "is any makefile or include out of
On Fri, Jul 4, 2014 at 8:55 PM, Paolo Bonzini wrote:
> Il 04/07/2014 14:27, Ming Lei ha scritto:
>
>> Now requests are submitted as a batch, so it is natural
>> to notify guest as a batch too.
>>
>> This may supress interrupt notification to VM:
>>
>> - in my test, decreased by ~13K/sec
>
I realize now that patch #5 and #6 of this series feel through the
cracks. Could you please rebase and resend?
On Mon, 27 Jan 2014, Wei Liu wrote:
> Signed-off-by: Wei Liu
> ---
> configure |7 +--
> default-configs/xenpv-softmmu.mak |2 ++
> 2 files changed,
Hi everybody,
We are experimenting with multi-core QEMU. We have Multiple QEMU cores
running
on multiple host threads and we are now looking at the issue of ‘atomic’
instructions.
Our initial thought was to push some sort of set of flags out to the memory
chain (eg: ATOMIC or NORMAL much as th
- Mail original -
> De: "Paolo Bonzini"
> À: "Sebastian Tanase" , qemu-devel@nongnu.org
> Cc: aligu...@amazon.com, afaer...@suse.de, r...@twiddle.net, "peter maydell"
> ,
> mich...@walle.cc, a...@alex.org.uk, stefa...@redhat.com,
> lcapitul...@redhat.com, crobi...@redhat.com,
> arm...@
Il 04/07/2014 16:33, Peter Maydell ha scritto:
On 30 June 2014 13:09, Paolo Bonzini wrote:
Il 28/06/2014 18:59, Peter Maydell ha scritto:
Having a direct dependency
Makefile: $(GENERATED_HEADERS)
can result in not-from-clean builds failing sometimes, because it means
that when Make does it
Il 04/07/2014 16:52, Ming Lei ha scritto:
> What you can do is change notify_guest to something like
>
> qemu_bh_schedule(req->dev->dataplane->notify_guest_bh);
>
> and do the actual notification in the bottom half. This should ensure that
> multiple notifications are coalesced, but it may a
Il 04/07/2014 17:36, Sebastian Tanase ha scritto:
I've also made some changes to patch 4 (added Sleep() function call when
compiling
for Windows and other small modifications); is it ok if I merge them with the
existing
patch 4 on your "icount" branch and then rebase the other patches?
Sure.
> > Thanks Dave:
> > Whether the randomness value/branch/code path the PVM and SVM
> may
> > have, It is only a performance issue. COLO never assumes the PVM and
> > SVM has same internal Machine state. From correctness p.o.v, as if
> > the PVM and SVM generate Identical response, we can view
Some callers use larger qiovs than nb_sectors for their read requests. This is
bad because raw-posix uses nb_sectors to allocate a bounce buffer and then
copies the whole iov over.
This series fixes some more cases (mirroring was already fixed earlier this
week) and adds assertions to catch any ot
At least raw-posix relies on this because it can allocate bounce buffers
based on the request length, but access it using all of the qiov entries
later.
Signed-off-by: Kevin Wolf
---
block.c | 2 ++
block/raw-posix.c | 15 +++
2 files changed, 13 insertions(+), 4 deletions
If a qcow2 image has a shorter backing file and a read request to
unallocated clusters goes across EOF of the backing file, the backing
file sees a shortened request and the rest is filled with zeros.
However, the original too long qiov was used with the shortened request.
This patch makes the qio
If a QED image has a shorter backing file and a read request to
unallocated clusters goes across EOF of the backing file, the backing
file sees a shortened request and the rest is filled with zeros.
However, the original too long qiov was used with the shortened request.
This patch makes the qiov
If a read request goes across EOF, the block driver sees a shortened
request that stops at EOF (the rest is memsetted in block.c), however
the original qiov was used for this request.
This patch makes the qiov size match the request size, avoiding a
potential buffer overflow in raw-posix.
Signed-
On Fri, Jul 4, 2014 at 11:48 PM, Paolo Bonzini wrote:
> Il 04/07/2014 16:52, Ming Lei ha scritto:
>
>>> > What you can do is change notify_guest to something like
>>> >
>>> > qemu_bh_schedule(req->dev->dataplane->notify_guest_bh);
>>> >
>>> > and do the actual notification in the bottom half.
On 4 July 2014 16:35, Frederic Konrad wrote:
> Hi everybody,
>
> We are experimenting with multi-core QEMU. We have Multiple QEMU cores
> running
> on multiple host threads and we are now looking at the issue of ‘atomic’
> instructions.
>
> Our initial thought was to push some sort of set of flags
Can we pretty please take me off the list of recepients. I have nothing to do
with this, I receive all list emails already, and where I am now it is quite
dificult to sort mail, and it costs me quite some time to do so too.
When I'm back I'll concider dropping mails sent to list and Cc'd to me.
Il 04/07/2014 17:57, Ming Lei ha scritto:
But we have two cases to consider:
- one submitted IO includes requests from multi vq(virtio-blk or
virtio-scsi maybe),
and each vq has to notify guest
- one submitted IO includes requests from multi bs for scsi device
The 2nd case should be easy to ha
From: "Dr. David Alan Gilbert"
Hi,
This is the 1st cut of my version of postcopy; it is designed for use with
the Linux kernel additions recently posted by Andrea Arcangeli here:
http://lists.gnu.org/archive/html/qemu-devel/2014-07/msg00525.html
The current status is:
1) It works - I've
From: "Dr. David Alan Gilbert"
The PMI holds the state of each page on the incoming side,
so that we can tell if the page is missing, already received
or there is a request outstanding for it.
Signed-off-by: Dr. David Alan Gilbert
---
include/migration/migration.h| 18 ++
include/migr
From: "Dr. David Alan Gilbert"
Stefan Berger's patch to create a QEMUFile that goes to a memory buffer;
from:
http://lists.gnu.org/archive/html/qemu-devel/2013-03/msg05036.html
Using the QEMUFile interface, this patch adds support functions for
operating on in-memory sized buffers that can be w
From: "Dr. David Alan Gilbert"
Improve the existing DPRINTF macros in migration.c and arch_init
by:
1) Making them go to stderr rather than stdout (so you can run with
-nographic and redirect your debug to a file)
2) Making them print the ms time with each debug - useful for
debugging latency
From: "Dr. David Alan Gilbert"
Postcopy needs a method to send messages from the destination back to
the source, this is the 'return path'.
Wire it up for 'socket' QEMUFile's using a dup'd fd.
Signed-off-by: Dr. David Alan Gilbert
---
include/migration/qemu-file.h | 8 +
qemu-file.c
From: "Dr. David Alan Gilbert"
Signed-off-by: Dr. David Alan Gilbert
---
arch_init.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch_init.c b/arch_init.c
index c006d21..58eccc1 100644
--- a/arch_init.c
+++ b/arch_init.c
@@ -439,6 +439,7 @@ ram_addr_t
migration_bitmap_find_and_reset_di
From: "Dr. David Alan Gilbert"
The return path uses a non-blocking fd so as not to block waiting
for the (possibly broken) destination to finish returning a message,
however we still want outbound data to behave in the same way and block.
Signed-off-by: Dr. David Alan Gilbert
---
qemu-file.c |
From: "Dr. David Alan Gilbert"
Create QEMU_VM_COMMAND section type for sending commands from
source to destination. These commands are not intended to convey
guest state but to control the migration process.
For use in postcopy.
Signed-off-by: Dr. David Alan Gilbert
---
include/migration/mig
From: "Dr. David Alan Gilbert"
Add two src->dest commands:
* OPENRP - To request that the destination open the return path
* REQACK - Request an acknowledge from the destination
Signed-off-by: Dr. David Alan Gilbert
---
include/migration/migration.h | 2 ++
include/sysemu/sysemu.h
From: "Dr. David Alan Gilbert"
Signed-off-by: Dr. David Alan Gilbert
---
docs/migration.txt | 148 +
1 file changed, 148 insertions(+)
diff --git a/docs/migration.txt b/docs/migration.txt
index 0492a45..dbd5e5f 100644
--- a/docs/migration.txt
From: "Dr. David Alan Gilbert"
userfaultfd is a Linux syscall that gives an fd that receives a stream
of notifications of accesses to pages marked as MADV_USERFAULT, and
allows the program to acknowledge those stalls and tell the accessing
thread to carry on.
Signed-off-by: Dr. David Alan Gilber
From: "Dr. David Alan Gilbert"
'MIG_STATE_POSTCOPY_ACTIVE' is entered after the precopy timelimit
has expired and migration switches to postcopy.
'migration_postcopy_phase' is provided for other sections to know if
they're in postcopy.
Signed-off-by: Dr. David Alan Gilbert
---
include/migrati
From: "Dr. David Alan Gilbert"
Tweak the end of migration cleanup; we don't want to close stuff down
at the end of the main stream, since the postcopy is still sending pages
on the other thread.
Signed-off-by: Dr. David Alan Gilbert
---
migration.c | 20
1 file changed, 20
From: "Dr. David Alan Gilbert"
Where postcopy is preceeded by a period of precopy, the destination will
have received pages that may have been dirtied on the source after the
page was sent. The destination must throw these pages away before
starting it's CPUs.
Maintain a 'sentmap' of pages that
From: "Dr. David Alan Gilbert"
When postcopy calls qemu_savevm_state_complete it's not really
the end of migration, so skip:
a) Finishing postcopiable iterative devices - they'll carry on
b) The termination byte on the end of the stream.
We then also add:
qemu_savevm_state_postcopy_compl
From: "Dr. David Alan Gilbert"
Provide functions to be called before the start of a postcopy
enabled migration (even if it's not eventually used) and
at the end.
During the init we must disable huge pages in the RAM that
we will receive postcopy data into, since if they start off
as hugepage and
From: "Dr. David Alan Gilbert"
Signed-off-by: Dr. David Alan Gilbert
---
arch_init.c | 11 +++
include/migration/migration.h | 1 +
migration.c | 2 ++
3 files changed, 14 insertions(+)
diff --git a/arch_init.c b/arch_init.c
index 134ea7e..fd7399c
From: "Dr. David Alan Gilbert"
Signed-off-by: Dr. David Alan Gilbert
---
include/migration/postcopy-ram.h | 5 +
postcopy-ram.c | 36 +++-
2 files changed, 40 insertions(+), 1 deletion(-)
diff --git a/include/migration/postcopy-ram.h b/inc
From: "Dr. David Alan Gilbert"
When transmitting RAM pages, consume pages that have been queued by
MIG_RPCOMM_REQPAGE commands and send them ahead of normal page scanning.
Note:
a) After a queued page the linear walk carries on from after the
unqueued page; there is a reasonable chance that th
From: "Dr. David Alan Gilbert"
postcopy_start:
Perform all the initialisation associated with starting up postcopy
mode from the source.
Signed-off-by: Dr. David Alan Gilbert
---
migration.c | 85 +
1 file changed, 85 insertions(+)
From: "Dr. David Alan Gilbert"
Switch to postcopy if:
1) There's still a significant amount to transfer
2) Postcopy is enabled
3) It's taken longer than the time set by the parameter.
and change the cleanup at the end of migration to match.
Signed-off-by: Dr. David Alan Gilbert
---
m
From: "Dr. David Alan Gilbert"
Signed-off-by: Dr. David Alan Gilbert
---
migration.c | 12
1 file changed, 12 insertions(+)
diff --git a/migration.c b/migration.c
index c73fcfa..c7ba6a1 100644
--- a/migration.c
+++ b/migration.c
@@ -1164,6 +1164,18 @@ void migrate_fd_connect(Migra
From: "Dr. David Alan Gilbert"
Signed-off-by: Dr. David Alan Gilbert
---
include/migration/migration.h | 3 +++
postcopy-ram.c| 23 +++
2 files changed, 26 insertions(+)
diff --git a/include/migration/migration.h b/include/migration/migration.h
index 397f41
1 - 100 of 172 matches
Mail list logo
