pygrub and qemuu need to be able to access a VM's disks locally in
order to be able to pull out the kernel and provide emulated disk
access, respectively. This can be done either by accessing the local
disk directly, or by plugging the target disk into dom0 to allow
access.
Unfortunately, while the plugging machinery works for pygrub, it does
not yet work for qemuu; this means that at the moment, disks with
hotplug scripts or disks with non-dom0 backends cannot be provided as
emulated devices to HVM domains.
Fortunately, disks using hotplug scripts created in dom0 do create a
block device as part of set-up, which can be accessed locally; and if
they use block-common.sh:write_dev, this path will be written to
physical-device-path.
Modify libxl__device_disk_setdefault() to be able to fish this path
out of xenstore and pass it to the caller.
Unfortunately, at the time pygrub runs, the devices have not yet been
set up. Rather than try to stash the domid somewhere to pass, we just
pass INVALID_DOMID.
This allows qemuu to emulate block devices created with custom hotplug
scripts.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Tested-by: George Dunlap <george.dunlap@citrix.com>
qemu can also access disks which will be provided with a qdisk backend
directly; add a flag to libxl__device_disk_find_local_path to indicate
whether to check for qdisk direct access.
Call libxl__device_disk_find_local_path() for most paths. If we can't
find a local path, print an error and skip the disk, rather than using
a bogus path.
Now if there is no local access to the disk (i.e., because the disk
has a non-local backend, or relies on a custom hotplug script), libxl
will now print a warning and not provide the emulated disk, rather
than providing bogus parameters to qemu which cause it to error out.
(Such disks will still be available via the PV backend.)
I left the libxl__blktap_devpath in the qemuu-specific code rather
than sharing it with the pyrgub code because:
1) When the pygrub path runs the guest disks have not yet been set up
2) libxl__blktap_devpath() will give you the existing devpath if it
already exists, but will set one up for you if you don't. So on the
pygrub path, this would end up setting up a new tap device.
3) There is no tap-specific teardown code on the pygrub path, and I
don't want to add any (particularly since I'm hoping to remove tapdisk
altogether soon).
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Tested-by: George Dunlap <george.dunlap@citrix.com>
Reorganize the qemuu disk argument code to make a clean separation
between finding a file to use, and constructing the parameters:
* Rename pdev_path to target_path
* Only use qemu_disk_format_string() in circumstances where qemu may
be interpreting the disk (i.e., backend==QDISK). In all other cases,
it should use RAW.
* Share as much as possible between the is_cdrom path and the normal
path.
This is mainly prep for sharing the local path finder with the
bootloader; but it does allow cdroms to use any backend that a normal
disk can use. Previously this was limited to RAW files or things that
qemu could handle directly; as of this changeset, it now includes tap
disks; and in future changesets it will include backends with custom
block scripts.
NB that this retains an existing bug, that disks with custom block
scripts or non-dom0 backends will have the bogus pdev_path passed in
to qemu, most likely resulting in qemu exiting with an error. This
will be fixed in follow-up patches.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Tested-by: George Dunlap <george.dunlap@citrix.com>
Move pygrub checks for local access ability into a separate function.
Also reorganize libxl__device_disk_local_initiate_attach so that we
don't initialize dls->disk unless we actually end up doing a local
attach.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Tested-by: George Dunlap <george.dunlap@citrix.com>
Change block-common.sh on Linux to write physical-device-path with the
path of the device node, in addition to physical-device with its
major:minor numbers.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Tested-by: George Dunlap <george.dunlap@citrix.com>
Regardless of whether we're running a custom hotplug script or using
normal phy: or file:, the "block" script will be run, which will set
all the necessary xenstore nodes.
In fact, writing this value here prevents the block script from
accomplishing its only purpose: to detect duplicate physical block
devices used in different virtual devices. The first thing the block
script does is check to see if this node is written; and if it is, it
silently exits.
Remove this, and let the block script perform its duplicate checking
function.
NOTE: It's likely that the duplicate checking for physical devices has
never been run under libxl (at least since this bug was introduced);
this may shake out some issues.
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Tested-by: George Dunlap <george.dunlap@citrix.com>
Testing the hotplug external script path at the moment involves
actually setting up one of the alternate datapaths (blktap, iscsi,
&c). Simplify testing by making a script which does a simple loopback,
but still has a target that can't be used directly.
To use:
script=block-dummy,vdev=xvda,target=dummy:<file>
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Tested-by: George Dunlap <george.dunlap@citrix.com>
During the investigation of very slow dump times of guest images in
Amazon EC2 instance, it was discovered that the
register_oldmem_pfn_is_ram() API implemented by the upstream kernel
commit 997c136f518c5debd63847e78e2a8694f56dcf90:
fs/proc/vmcore.c: add hook to read_from_oldmem() to check
for non-ram pages
was not being called. This was due to the PV driver with the call
to register_oldmem_pfn_is_ram() API was not including the
kernel header file that is used to communicate support of the API in the
kernel. Fix the issue by including the required header file.
Signed-off-by: Mike Meyer <mike.meyer@teradata.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Olaf Hering <olaf@aepfle.de>
Commit 108788e8 "x86/hvm: separate ioreq server code from generic hvm
code" split the code supporting ioreq servers from the rest of the HVM
code. Now that this is separate, the new files can be added to the x86
I/O emulation file list in MAINTAINERS.
Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
COLO introduced a few callbacks. The original implementation used
unsigned long for a type which in fact should be xen_pfn_t. That broke
libxl compilation on ARM, because xen_pfn_t is not a synonym for
unsigned long on ARM platform.
Fixing this requires modifying the perl script: specifically now we
need to include xenctrl.h before _libxl_save_msgs_*.h, rather than
afterwards, so that we can use xen_pfn_t there.
Reported-by: Olaf Hering <olaf@aepfle.de>
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
The call site is enclosed by x86 define guards.
Without this patch:
[ 334s] xc_resume.c:112:12: error: 'xc_domain_resume_hvm' defined but not used [-Werror=unused-function]
[ 334s] static int xc_domain_resume_hvm(xc_interface *xch, uint32_t domid)
Reported-by: Olaf Hering <olaf@aepfle.de>
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
The patches adding Xen tools support for paravirtualized USB devices
(pvUSB) omitted documenting the introduced Xenstore paths.
Add the paths to docs/misc/xenstore-paths.markdown
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
The hyperlinks for the different I/O protocols are wrong.
Correct them.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
The code in hvm/hvm.c related to handling I/O emulation using the ioreq
server framework is large and mostly self-contained.
This patch separates the ioreq server code into a new hvm/ioreq.c source
module and accompanying asm-x86/hvm/ioreq.h header file. There is no
intended functional change, only code movement.
Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Acked-by: Kevin Tian <kevin.tian@intel.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Add XEN_DOMCTL_SCHEDOP_getvcpuinfo and _putvcpuinfo hypercalls
to independently get and set the scheduling parameters of each
vCPU of a domain.
Also fix a bug in XEN_DOMCTL_SCHEDOP_getinfo, where PERIOD and
BUDGET are not divided by MICROSECS(1) before being retruned
to the caller.
Signed-off-by: Chong Li <chong.li@wustl.edu>
Signed-off-by: Meng Xu <mengxu@cis.upenn.edu>
Signed-off-by: Sisu Xi <xisisu@gmail.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Dario Faggioli <dario.faggioli@citrix.com>
This was introduced by 97ee1f (~5 years ago), but was probably never
surfaced because most people used regular files as CDROM images, so the PHY
backend was actually never selected. A year ago this was changed, and now
regular RAW files are also handled by the PHY backend, which has made this
bug suface.
Fix it by allowing empty disks to use the PHY backend, skipping the stat
tests.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reported-by: Alex Braunegg <alex.braunegg@gmail.com>
Current implementation of elf_load_bsdsyms is broken when loading inside of
a HVM guest, because it assumes elf_memcpy_safe is able to write into guest
memory space, which it is not.
Take the oportunity to do some cleanup and properly document how
elf_{parse/load}_bsdsyms works. The new implementation uses elf_load_image
when dealing with data that needs to be copied to the guest memory space.
Also reduce the number of section headers copied to the minimum necessary.
This patch also removes the duplication of code found in the libxc ELF
loader, since the libelf symtab/strtab loading code will also handle this
case without having to duplicate it.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
And use it as the default value for the VGA kind. This allows libxl to set
it to the default value later on when the domain type is known. For HVM
guests the default value is LIBXL_VGA_INTERFACE_TYPE_CIRRUS while for
HVMlite the default value is LIBXL_VGA_INTERFACE_TYPE_NONE.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
(i.e., domain creation and destruction) so the
trace will show properly decoded info, rather
than just a bunch of hex codes.
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Acked-by: George Dunlap <george.dunlap@citrix.com>
Before this patch:
1. Suspend
a. PVHVM and PV: we use the same way to suspend the guest (send the
suspend request to the guest). If the guest doesn't support evtchn, the
xenstore variant will be used, suspending the guest via XenBus control
node.
b. Pure HVM: we call xc_domain_shutdown(..., SHUTDOWN_suspend) to suspend
the guest.
2. Resume:
a. Fast path (fast=1)
Do not change the guest state. We call libxl__domain_resume(.., 1) which
calls xc_domain_resume(..., 1 /* fast=1*/) to resume the guest.
PV: Modify the return code to 1, and than call the domctl:
XEN_DOMCTL_resumedomain
PVHVM: same with PV
Pure HVM: Do nothing in modify_returncode, and than call the domctl:
XEN_DOMCTL_resumedomain
b. Slow path (fast=0)
Used when the guest's state have been changed. Will call
libxl__domain_resume(..., 0) to resume the guest.
PV: Update start info, and reset all secondary CPU states. Than call
the domctl: XEN_DOMCTL_resumedomain
PVHVM: Can not be resumed. You will get the following error message:
"Cannot resume uncooperative HVM guests"
Pure HVM: Same with PVHVM
After this patch:
1. Suspend
Unchanged.
2. Resume
a. Fast path
Unchanged.
b. Slow path
PV: Unchanged.
PVHVM: Call XEN_DOMCTL_resumedomain to resume the guest. Because we
don't modify the return code, PV drivers will disconnect
and reconnect.
The guest ends up doing the XENMAPSPACE_shared_info
XENMEM_add_to_physmap hypercall and resetting all of its CPU
states to point to the shared_info (except the ones past 32).
That is the Linux kernel does that - regardless whether the
SCHEDOP_shutdown:SHUTDOWN_suspend returns 1 or not.
Pure HVM: Call XEN_DOMCTL_resumedomain to resume the guest.
Under COLO, we will update the guest's state (modify memory, CPU registers,
device state etc). In this case, we cannot use the fast path to resume it.
Keep the return code 0, and use the slow path to resume the guest.
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
[ wei: reformat commit message a bit ]
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Add cmdline switches to 'xl migrate-receive' command to specify
a domain-specific hotplug script to setup COLO proxy.
Add a new config var 'colo.default.agentscript' to xl.conf, that
allows the user to override the default global script used to
setup COLO proxy.
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Use qemu block replication as our block replication solution.
Note that guest must be paused before starting COLO, otherwise,
the disk won't be consistent between primary and secondary.
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
We will use qemu block replication, and qemu provides some qmp commands
to prepare replication, start replication, get replication error, and
stop replication. Introduce new API to execute these qmp commands.
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Add a new option -c to the command 'xl remus'. If you want
to use COLO HA instead of Remus HA, please use -c option.
Update man pages to reflect the addition of a new option to
'xl remus' command.
Also add a new option --colo to the internal command 'xl migrate-receive'.
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
a. call callbacks resume/checkpoint/suspend while secondary vm
status is consistent with primary
b. send dirty pfn list to primary when checkpoint under colo
c. send store gfn and console gfn to xl before resuming secondary vm
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
CC: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
We will do the following things again and again:
1. Suspend primary vm
a. Suspend primary vm
b. do postsuspend
c. Read CHECKPOINT_SVM_SUSPENDED sent by secondary
2. Checkpoint
a. Write emulator xenstore data and emulator context
b. Write checkpoint end record
3. Resume primary vm
a. Read CHECKPOINT_SVM_READY from slave
b. Do presume
c. Resume primary vm
d. Read CHECKPOINT_SVM_RESUMED from slave
4. Wait a new checkpoint
a. Wait a new checkpoint(not implemented)
b. Send CHECKPOINT_NEW to slave
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
No functional changes and this cleanup will make the later
patch called "primary vm suspend/resume/checkpoint code" not
too complicated.
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Secondary vm is running in colo mode. So we will do
the following things again and again:
1. Resume secondary vm
a. Send CHECKPOINT_SVM_READY to master.
b. If it is not the first resume, call libxl__checkpoint_devices_preresume().
c. If it is the first resume(resume right after live migration),
- call libxl__xc_domain_restore_done() to build the secondary vm.
- enable secondary vm's logdirty.
- call libxl__domain_resume() to resume secondary vm.
- call libxl__checkpoint_devices_setup() to setup checkpoint devices.
d. Send CHECKPOINT_SVM_RESUMED to master.
2. Wait a new checkpoint
a. Call libxl__checkpoint_devices_commit().
b. Read CHECKPOINT_NEW from master.
3. Suspend secondary vm
a. Suspend secondary vm.
b. Call libxl__checkpoint_devices_postsuspend().
c. Send CHECKPOINT_SVM_SUSPENDED to master.
4. Checkpoint
a. Read emulator xenstore data and emulator context
b. REC_TYPE_CHECKPOINT_END
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
This is used by primay to read records sent by secondary.
Note: The function libxl__stream_read_checkpoint_state() will be used
in later patches called "secondary vm suspend/resume/checkpoint code" and
"primary vm suspend/resume/checkpoint code".
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Add back channel support to write stream. If the write stream is
a back channel stream, this means the write stream is used by
Secondary to send some records back.
Note: The function libxl__stream_write_checkpoint_state() will be used
in later patches called "secondary vm suspend/resume/checkpoint code" and
"primary vm suspend/resume/checkpoint code".
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
read_record() could be used by primary to read dirty bitmap
record sent by secondary under COLO.
When used by xc save side, we need to pass the backchannel fd
instead of ctx->fd to read_record(), so we added a fd param to
it.
No functional changes.
CC: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
In COLO mode, both VMs are running, and are considered in sync if the
visible network traffic is identical. After some time, they fall out of
sync.
At this point, the two VMs have definitely diverged. Lets call the
primary dirty bitmap set A, while the secondary dirty bitmap set B.
Sets A and B are different.
Under normal migration, the page data for set A will be sent from the
primary to the secondary.
However, the set difference B - A (the one in B but not in A, lets
call this C) is out-of-date on the secondary (with respect to the
primary) and will not be sent by the primary (to secondary), as it
was not memory dirtied by the primary. The secondary needs C page data
to reconstruct an exact copy of the primary at the checkpoint.
The secondary cannot calculate C as it doesn't know A. Instead, the
secondary must send B to the primary, at which point the primary
calculates the union of A and B (lets call this D) which is all the
pages dirtied by both the primary and the secondary, and sends all page
data covered by D.
In the general case, D is a superset of both A and B. Without the
backchannel dirty bitmap, a COLO checkpoint can't reconstruct a valid
copy of the primary.
We transfer the dirty bitmap on libxc side, so we need to introduce back
channel to libxc.
Note: it is different from the paper. We change the original design to
the current one, according to our following concerns:
1. The original design needs extra memory on Secondary host. When there's
multiple backups on one host, the memory cost is high.
2. The memory cache code will be another 1k+, it will make the review
more time consuming.
Note: this patch merely adds new parameters to various prototypes and
functions. The new parameters are used in later patch called
"libxc/restore: send dirty pfn list to primary when checkpoint under
COLO".
Signed-off-by: Yang Hongyang <hongyang.yang@easystack.cn>
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
CC: Ian Campbell <Ian.Campbell@citrix.com>
CC: Ian Jackson <Ian.Jackson@eu.citrix.com>
CC: Wei Liu <wei.liu2@citrix.com>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
George Dunlap
George Dunlap
Mike Meyer
Paul Durrant
Ian Jackson
Wei Liu
Juergen Gross
Chong Li
Roger Pau Monne
Dario Faggioli
Olaf Hering
Konrad Rzeszutek Wilk
Wen Congyang
Changlong Xie