If a migration fails/is cancelled during the postcopy stage we currently
end up with the runstate as finish-migrate, where it should be post-migrate.
There's a small window in precopy where I think the same thing can
happen, but I've never seen it.
It rarely matters; the only postcopy case is if you restart a migration, which
again is a case that rarely matters in postcopy because it's only
safe to restart the migration if you know the destination hasn't
been running (which you might if you started the destination with -S
and hadn't got around to 'c' ing it before the postcopy failed).
Even then it's a small window but potentially you could hit if
there's a problem loading the devices on the destination.
This corresponds to:
https://bugzilla.redhat.com/show_bug.cgi?id=1355683
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Message-Id: <1468601086-32117-1-git-send-email-dgilbert@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
virgl conditionally registers a vmstate as unmigratable when virgl
is enabled; instead use the migrate_add_blocker mechanism.
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>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Forcibly convert it to a vmstate wrapper; proper conversion
comes later.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Forcibly convert it to a vmstate wrapper; proper conversion
comes later.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Forcibly convert it to a vmstate wrapper; proper conversion
comes later.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Forcibly convert it to a vmstate wrapper; proper conversion
comes later.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Forcibly convert it to a vmstate wrapper; proper conversion
comes later.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Forcibly convert it to a vmstate wrapper; proper conversion
comes later.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Forcibly convert it to a vmstate wrapper; proper conversion
comes later.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
To make conversion of virtio devices to VMState simple
at first add a helper function for the simple virtio_save
case and a helper macro that defines the VMState structure.
These will probably go away or change as more of the virtio
code gets converted.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
virtio-net has had version 11 since 0ce0e8f4 in 2009
(v0.11.0-rc0-1480-g0ce0e8f) - remove the code to support loading
anything earlier.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Add the host-phys-bits boolean property, if true, take phys-bits
from the hosts physical bits value, overriding either the default
or the user specified value.
We can also use the value we read from the host to check the users
explicitly set value and warn them if it doesn't match.
Note:
a) We only read the hosts value in KVM mode (because on non-x86
we get an abort if we try)
b) We don't warn about trying to use host-phys-bits in TCG mode,
we just fall back to the TCG default. This allows the machine
type to set the host-phys-bits flag if it wants and then to
work in both TCG and KVM.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Fill the bits between 51..number-of-physical-address-bits in the
MTRR_PHYSMASKn variable range mtrr masks so that they're consistent
in the migration stream irrespective of the physical address space
of the source VM in a migration.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
The CPU GPs if we try and set a bit in a variable MTRR mask above
the limit of physical address bits on the host. We hit this
when loading a migration from a host with a larger physical
address limit than our destination (e.g. a Xeon->i7 of same
generation) but previously used to get away with it
until 48e1a45 started checking that msr writes actually worked.
It seems in our case the GP probably comes from KVM emulating
that GP.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Currently QEMU sets the x86 number of physical address bits to the
magic number 40. This is only correct on some small AMD systems;
Intel systems tend to have 36, 39, 46 bits, and large AMD systems
tend to have 48.
Having the value different from your actual hardware is detectable
by the guest and in principal can cause problems;
The current limit of 40 stops TB VMs being created by those lucky
enough to have that much.
This patch lets you set the physical bits by a cpu property but
defaults to the same 40bits which matches TCGs setup.
I've removed the ancient warning about the 42 bit limit in exec.c;
I can't find that limit in there and no one else seems to know where
it is.
We use a magic value of 0 as the property default so that we can
later distinguish between the default and a user set value.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
A couple of improvements to tracing that have come out of helping
people with migration problems:
* vmstate_n_elems trace the count/name - for when you have problems
getting array counts right
* vmstate_subsection_load_bad - add the idstr, for when you receive a
subsection you weren't expecting.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-Id: <1465896986-16132-1-git-send-email-dgilbert@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Use the avx2 primitives during the test, thus making sure that the
compiler and assembler could actually use avx2.
This also detects the failure case on gcc 4.8.x with -save-temps
and avoids the need for the gcc version check in cutils.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-Id: <1465557378-24105-3-git-send-email-dgilbert@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This is a postcopy test (x86 only) that actually runs the guest
and checks the memory contents.
The test runs from an x86 boot block with the hex embedded in the test;
the source for this is:
...........
.code16
.org 0x7c00
.file "fill.s"
.text
.globl start
.type start, @function
start: # at 0x7c00 ?
cli
lgdt gdtdesc
mov $1,%eax
mov %eax,%cr0 # Protected mode enable
data32 ljmp $8,$0x7c20
.org 0x7c20
.code32
# A20 enable - not sure I actually need this
inb $0x92,%al
or $2,%al
outb %al, $0x92
# set up DS for the whole of RAM (needed on KVM)
mov $16,%eax
mov %eax,%ds
mov $65,%ax
mov $0x3f8,%dx
outb %al,%dx
# bl keeps a counter so we limit the output speed
mov $0, %bl
mainloop:
# Start from 1MB
mov $(1024*1024),%eax
innerloop:
incb (%eax)
add $4096,%eax
cmp $(100*1024*1024),%eax
jl innerloop
inc %bl
jnz mainloop
mov $66,%ax
mov $0x3f8,%dx
outb %al,%dx
jmp mainloop
# GDT magic from old (GPLv2) Grub startup.S
.p2align 2 /* force 4-byte alignment */
gdt:
.word 0, 0
.byte 0, 0, 0, 0
/* -- code segment --
* base = 0x00000000, limit = 0xFFFFF (4 KiB Granularity), present
* type = 32bit code execute/read, DPL = 0
*/
.word 0xFFFF, 0
.byte 0, 0x9A, 0xCF, 0
/* -- data segment --
* base = 0x00000000, limit 0xFFFFF (4 KiB Granularity), present
* type = 32 bit data read/write, DPL = 0
*/
.word 0xFFFF, 0
.byte 0, 0x92, 0xCF, 0
gdtdesc:
.word 0x27 /* limit */
.long gdt /* addr */
/* I'm a bootable disk */
.org 0x7dfe
.byte 0x55
.byte 0xAA
...........
and that can be assembled by the following magic:
as --32 -march=i486 fill.s -o fill.o
objcopy -O binary fill.o fill.boot
dd if=fill.boot of=bootsect bs=256 count=2 skip=124
xxd -i bootsect
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Message-id: 1465816605-29488-5-git-send-email-dgilbert@redhat.com
Message-Id: <1465816605-29488-5-git-send-email-dgilbert@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Pflash migration (e.g. q35 + EFI variable storage) fails
with the assert:
bdrv_co_do_pwritev: Assertion `!(bs->open_flags & 0x0800)' failed.
This avoids the problem by delaying the pflash update until after
the device loads complete.
Tested by:
Migrating Q35/EFI vm.
Changing efi variable content (with efiboot in the guest)
md5sum'ing the variable file before migration and after.
This is a fix that Paolo posted in the message
570244B3.4070105@redhat.com
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Acked-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The i82801b11 bridge didn't have a vmsd and thus didn't send
any migration data, including that of its parent PCIBridge object.
The symptom being if the guest used any devices behind the bridge
the guest crashed (mostly with various interrupt related issues).
Note: This will cause migration from old qemus that used this device to
explicitly fail during migration as opposed to the guest crashing.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Suggested-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Spice hooks the migration status changes to figure out when to
transmit information to the new spice server; but the migration
status in postcopy doesn't quite fit - the destination starts
running before the end of the source migration.
It's not a case of hanging off the migration status change to
postcopy-active either, since that happens before we stop the
guest CPU.
Fix it by sending a notify just after sending the device state,
and adding a flag that can be tested by the notify receiver.
Symptom:
spice handover doesn't work with the error:
red_worker.c:11540:display_channel_wait_for_migrate_data: timeout
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Message-id: 1456161452-25318-1-git-send-email-dgilbert@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
I misunderstood the vmstate macro definition when I reworked the
virtio .get/.put.
The VMSTATE_STRUCT_VARRAY_KNOWN, was described as being for "a
variable length array (i.e. _type *_field) but we know the
length". However it actually specified operation for arrays embedded in
the struct (i.e. _type _field[]) since it lacked the VMS_POINTER
flag. This caused offset calculation to be completely off, examining and
potentially sending random data instead of the VirtQueue content.
Replace the otherwise unused VMSTATE_STRUCT_VARRAY_KNOWN with a
VMSTATE_STRUCT_VARRAY_POINTER_KNOWN that includes the VMS_POINTER flag
(so now actually doing what it advertises) and use it in the virtio
migration code.
Fixes and description as per Sascha's suggestions/debug.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reported-by: Sascha Silbe <silbe@linux.vnet.ibm.com>
Tested-By: Sascha Silbe <silbe@linux.vnet.ibm.com>
Reviewed-By: Sascha Silbe <silbe@linux.vnet.ibm.com>
Fixes: 50e5ae4dc3
Fixes: 2cf0148674
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Jiri saw a hang on pause_all_vcpus called from postcopy_start,
where the cpus are all apparently stopped ('stopped' flag set)
but pause_all_vcpus is still stuck on a cond_wait on qemu_paused_cond.
We suspect this is happening if a qmp_stop is called at about the
same time as the postcopy code calls that pause_all_vcpus;
although they both should have the main lock held, Paolo spotted
the cond_wait unlocks the global lock so perhaps they both
could end up waiting at the same time?
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reported-by: Jiri Denemark <jdenemar@redhat.com>
Message-Id: <1453716498-27238-1-git-send-email-dgilbert@redhat.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
eth.h and slirp.h both define ETH_ALEN and ETH_P_IP
rtl8139.c and eth.h both define ETH_HLEN
Move the related constant (ETH_P_ARP) from slirp.h to eth.h, and
remove the duplicates; make slirp.h include eth.h
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Jason Wang <jasowang@redhat.com>
The 'virtqueue_state' and 'ringsize' can be saved using VMSTATE
macros rather than hand coded .get/.put
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>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
At the moment we have VMSTATE_STRUCT_ARRAY that requires
the field is declared as an array of fixed size.
We also have VMSTATE_STRUCT_VARRAY_UINT* that allows
a field declared as a pointer, but requires that the length
is a field member in the structure being loaded/saved.
VMSTATE_STRUCT_VARRAY_KNOWN is for arrays defined as pointers
yet we somehow know the length of.
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>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
My fix (84e7b80a) replaced the last_sent_block update that I'd
removed earlier; however it was too aggressive in the xbzrle case.
save_xbzrle_page might return '0' to mean that the page didn't
need sending since it was the same as the last sent version;
in this case we can't update 'last_sent_block' since we didn't
actually send it.
Symptom: 'Illegal RAM offset 1018000' as we try and send a page
to the wrong RAMBlock; potentially that could be a data
corruption if you were really unlucky.
Fixes: 84e7b80a05
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Message-id: 1449765106-6528-1-git-send-email-dgilbert@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
madvise() returns EINVAL in the case of many failures, but also
returns it in cases where the host kernel doesn't have THP enabled.
Postcopy only really cares that THP is off before it detects faults,
and turns it back on afterwards; so we're going to have
to assume that if the madvise fails then the host just doesn't do
THP and we can carry on with the postcopy.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Tested-by: Jason J. Herne <jjherne@linux.vnet.ibm.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The check is unneccesary, we read the value at the start of the
thread, use it, and never change it. The value is checked to be
non-NULL before thread creation.
Spotted by coverity, CID 1339211
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
I set current_time before the postcopy test but never use it;
(I think this was from the original version where it was time based).
Spotted by coverity, CID 1339208
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
In a82d593b61 I accidentally removed the setting of
last_sent_block, put it back.
Symptoms:
Multithreaded compression only uses one thread.
Migration is a bit less efficient since it won't use 'cont' flags.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Fixes: a82d593b61
Signed-off-by: Juan Quintela <quintela@redhat.com>
Peter reported a lock error on MacOS after my a82d593b
patch.
migrate_get_current does one-time initialisation of
a bunch of variables.
migrate_init does reinitialisation even on a 2nd
migrate after a cancel.
The problem here was that I'd initialised the mutex
in migrate_get_current, and the memset in migrate_init
corrupted it.
Remove the memset and replace it by explicit initialisation
of fields that need initialising; this also turns out to be simpler
than the old code that had to preserve some fields.
Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Fixes: a82d593b
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Improve the text in both the qapi-schema and hmp help to point out
you need to set the postcopy-ram capability prior to issuing
migrate-start-postcopy.
Also fix the text of the migrate_start_postcopy error that
deals with capabilities.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Acked-by: Jason J. Herne <jjherne@linux.vnet.ibm.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Where the target page size is different from the host page
we special case it, but I messed up on the zero case check.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Where we have iterable, but non-postcopiable devices (e.g. htab
or block migration), complete them before forming the 'package'
but with the CPUs stopped. This stops them filling up the package.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Postcopy detects accesses to pages that haven't been transferred yet
using userfaultfd, and it causes exceptions on pages that are 'not
present'.
Ballooning also causes pages to be marked as 'not present' when the
guest inflates the balloon.
Potentially a balloon could be inflated to discard pages that are
currently inflight during postcopy and that may be arriving at about
the same time.
To avoid this confusion, disable ballooning during postcopy.
When disabled we drop balloon requests from the guest. Since ballooning
is generally initiated by the host, the management system should avoid
initiating any balloon instructions to the guest during migration,
although it's not possible to know how long it would take a guest to
process a request made prior to the start of migration.
Guest initiated ballooning will not know if it's really freed a page
of host memory or not.
Queueing the requests until after migration would be nice, but is
non-trivial, since the set of inflate/deflate requests have to
be compared with the state of the page to know what the final
outcome is allowed to be.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Userfault doesn't work with mlock; mlock is designed to nail down pages
so they don't move, userfault is designed to tell you when they're not
there.
munlock the pages we userfault protect before postcopy.
mlock everything again at the end if mlock is enabled.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
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 <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Prior to servicing userfault requests we must ensure we've not got
huge pages in the area that might include non-transferred memory,
since a hugepage could incorrectly mark the whole huge page as present.
We mark the area as non-huge page (nhp) just before we perform
discards; the discard code now tells us to discard any areas
that haven't been sent (as well as any that are redirtied);
any already formed transparent-huge-pages get fragmented
by this discard process if they cotnain any discards.
Transparent huge pages that have been entirely transferred
and don't contain any discards are not broken by this mechanism;
they stay as huge pages.
By starting postcopy after a full precopy pass, many of the pages
then stay as huge pages; this is important for maintaining performance
after the end of the migration.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Wire up more of the handlers for the commands on the destination side,
in particular loadvm_postcopy_handle_run now has enough to start the
guest running.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The loading of a device state (during postcopy) may access guest
memory that's still on the source machine and thus might need
a page fill; split off a separate thread that handles the incoming
page data so that the original incoming migration code can finish
off the device data.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
userfaultfd is a Linux syscall that gives an fd that receives a stream
of notifications of accesses to pages registered with it and allows
the program to acknowledge those stalls and tell the accessing
thread to carry on.
We convert the requests from the kernel into messages back to the
source asking for the pages.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
RAMBlocks that are not a multiple of host pages in length
cause problems for postcopy (I've seen an ACPI table on aarch64
be 5k in length - i.e. 5x target-page), so round RAMBlock sizes
up to a host-page.
This potentially breaks migration compatibility due to changes
in RAMBlock sizes; however:
1) x86 and s390 I think always have host=target page size
2) When I've tried on Power the block sizes already seem aligned.
3) I don't think there's anything else that maintains per-version
machine-types for compatibility.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Prior to the start of postcopy, ensure that everything that will
be transferred later is a whole host-page in size.
This is accomplished by discarding partially transferred host pages
and marking any that are partially dirty as fully dirty.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
During the postcopy phase we must not call the iterate method on
precopy-only devices, since they may have done some cleanup during
the _complete call at the end of the precopy phase.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Once we're in postcopy the source processors are stopped and memory
shouldn't change any more, so there's no need to look at the dirty
map.
There are two notes to this:
1) If we do resync and a page had changed then the page would get
sent again, which the destination wouldn't allow (since it might
have also modified the page)
2) Before disabling this I'd seen very rare cases where a page had been
marked dirtied although the memory contents are apparently identical
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Ensure that target pages received within a host page are in order.
This shouldn't trigger, but in the cases where the sender goes
wrong and sends stuff out of order it produces a corruption that's
really nasty to debug.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
In postcopy, the destination guest is running at the same time
as it's receiving pages; as we receive new pages we must put
them into the guests address space atomically to avoid a running
CPU accessing a partially written page.
Use the helpers in postcopy-ram.c to map these pages.
qemu_get_buffer_in_place is used to avoid a copy out of qemu_file
in the case that postcopy is going to do a copy anyway.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
postcopy_place_page (etc) provide a way for postcopy to place a page
into guests memory atomically (using the copy ioctl on the ufd).
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
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 the destination
was about to ask for other closeby pages anyway.
b) We have to be careful of any assumptions that the page walking
code makes, in particular it does some short cuts on its first linear
walk that break as soon as we do a queued page.
c) We have to be careful to not break up host-page size chunks, since
this makes it harder to place the pages on the destination.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Add MIG_RP_MSG_REQ_PAGES command on Return path for the postcopy
destination to request a page from the source.
Two versions exist:
MIG_RP_MSG_REQ_PAGES_ID that includes a RAMBlock name and start/len
MIG_RP_MSG_REQ_PAGES that just has start/len for use with the same
RAMBlock as a previous MIG_RP_MSG_REQ_PAGES_ID
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The end of migration in postcopy is a bit different since some of
the things normally done at the end of migration have already been
done on the transition to postcopy.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Soon we'll be in either ACTIVE or POSTCOPY_ACTIVE when we
complete migration, and we need to know which we expect to be
in to change state safely.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Maintain an 'unsentmap' of pages that have yet to be sent.
This is used in the following patches to discard some set of
the pages already sent as we enter postcopy mode.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Add qemu_savevm_state_complete_postcopy to complement
qemu_savevm_state_complete_precopy together with a new
save_live_complete_postcopy method on devices.
The save_live_complete_precopy method is called on
all devices during a precopy migration, and all non-postcopy
devices during a postcopy migration at the transition.
The save_live_complete_postcopy method is called at
the end of postcopy for all postcopiable devices.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
VMDescription is normally sent at the end, after all
of the devices; however that's not the end for postcopy,
so just don't send it when in postcopy.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Soon we'll be in either ACTIVE or POSTCOPY_ACTIVE when we
complete migration, and we need to know which we expect to be
in to change state safely.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Once postcopy is enabled (with migrate_set_capability), the migration
will still start on precopy mode. To cause a transition into postcopy
the:
migrate_start_postcopy
command must be issued. Postcopy will start sometime after this
(when it's next checked in the migration loop).
Issuing the command before migration has started will error,
and issuing after it has finished is ignored.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Provide a check to see if the OS we're running on has all the bits
needed for postcopy.
Creates postcopy-ram.c which will get most of the other helpers we need.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
MIG_CMD_PACKAGED is a migration command that wraps a chunk of migration
stream inside a package whose length can be determined purely by reading
its header. The destination guarantees that the whole MIG_CMD_PACKAGED
is read off the stream prior to parsing the contents.
This is used by postcopy to load device state (from the package)
while leaving the main stream free to receive memory pages.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The state of the postcopy process is managed via a series of messages;
* Add wrappers and handlers for sending/receiving these messages
* Add state variable that track the current state of postcopy
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The 'postcopy ram' capability allows postcopy migration of RAM;
note that the migration starts off in precopy mode until
postcopy mode is triggered (see the migrate_start_postcopy
patch later in the series).
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Postcopy needs to have two migration streams loading concurrently;
one from memory (with the device state) and the other from the fd
with the memory transactions.
Split the core of qemu_loadvm_state out so we can use it for both.
Allow the inner loadvm loop to quit and cause the parent loops to
exit as well.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Add 'migration_is_setup_or_active' utility function to check state.
(It gets postcopy added to it's list later on in the series)
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Add migrate_send_rp_message to send a message from destination to source along the return path.
(It uses a mutex to let it be called from multiple threads)
Add migrate_send_rp_shut to send a 'shut' message to indicate
the destination is finished with the RP.
Add migrate_send_rp_ack to send a 'PONG' message in response to a PING
Use it in the MSG_RP_PING handler
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Add two src->dest commands:
* OPEN_RETURN_PATH - To request that the destination open the return path
* PING - Request an acknowledge from the destination
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
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 <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The destination sets the fd to non-blocking on incoming migrations;
this also affects the return path from the destination, and thus we
need to make sure we can safely write to the return path.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
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.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Postcopy uses the userfaultfd.h feature in the Linux kernel; include
the header.
(In early versions of the patch series we had this, and then we dropped
this by only including it if the kernel headers defined the syscall
number; however 1842bdfd added the syscall definition to our
headers, which means we can't tell if the kernel has it or not)
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
In postcopy we're going to need to perform the complete phase
for postcopiable devices at a different point, start out by
renaming all of the 'complete's to make the difference obvious.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Suspend to file is very much like a migrate, and it makes life
easier if we have the Migration state available, so initialise it
in the savevm.c code for suspending.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewd-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
The main RAM load loop has a call to host_from_stream_offset for
each page type that actually loads data with the same test;
factor it out before the switch.
The host = NULL is to silence a bogus gcc warning of
an unitialised in the RAM_SAVE_COMPRESS_PAGE case, it
doesn't seem to realise that host is always initialised by the if at
the top in the cases the switch takes.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Useful for debugging the migration bitmap and other bitmaps
of the same format (including the sentmap in postcopy).
The bitmap is printed to stderr.
Lines that are all the expected value are excluded so the output
can be quite compact for many bitmaps.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
Add QEMU_MADV_NOHUGEPAGE as an OS-independent version of
MADV_NOHUGEPAGE.
We include sys/mman.h before making the test to ensure
that we pick up the system defines.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
