# gpg: Signature made Thu 18 May 2017 06:23:26 PM BST
# gpg: using RSA key 0xF487EF185872D723
# gpg: Good signature from "Juan Quintela <[email protected]>"
# gpg: aka "Juan Quintela <[email protected]>"
# Primary key fingerprint: 1899 FF8E DEBF 58CC EE03 4B82 F487 EF18 5872 D723
* quintela/tags/migration/20170518:
migration: Make savevm.c target independent
exec: Create include for target_page_size()
migration: migration.h was not needed
migration: Remove vmstate.h from migration.h
migration: Remove qemu-file.h from vmstate.h
migration: Split vmstate-types.c from vmstate.c
migration: Move qjson.h to migration/
migration: Remove migration.h from colo.h
migration: Export qemu-file-channel.c functions in its own file
migration: Split migration/channel.c for channel operations
migration: Create migration/xbzrle.h
block migration: Allow compile time disable
migration: Remove old MigrationParams
migration: Remove use of old MigrationParams
migration: Create block capability
hmp: Use visitor api for hmp_migrate_set_parameter()
postcopy: Require RAMBlocks that are whole pages
migration: Fix non-multiple of page size migration
Dong Jia Shi [Wed, 17 May 2017 00:48:12 +0000 (02:48 +0200)]
vfio/ccw: update sense data if a unit check is pending
Concurrent-sense data is currently not delivered. This patch stores
the concurrent-sense data to the subchannel if a unit check is pending
and the concurrent-sense bit is enabled. Then a TSCH can retreive the
right IRB data back to the guest.
Implement a basic infrastructure of handling channel I/O instruction
interception for passed through subchannels:
1. Branch the code path of instruction interception handling by
SubChannel type.
2. For a passed-through subchannel, issue the ORB to kernel to do ccw
translation and perform an I/O operation.
3. Assign different condition code based on the I/O result, or
trigger a program check.
s390x/css: introduce and realize ccw-request callback
Introduce a new callback on subchannel to handle ccw-request.
Realize the callback in vfio-ccw device. Besides, resort to
the event notifier handler to handling the ccw-request results.
1. Pread the I/O results via MMIO region.
2. Update the scsw info to guest.
3. Inject an I/O interrupt to notify guest the I/O result.
Dong Jia Shi [Wed, 17 May 2017 00:48:09 +0000 (02:48 +0200)]
vfio/ccw: get irqs info and set the eventfd fd
vfio-ccw resorts to the eventfd mechanism to communicate with userspace.
We fetch the irqs info via the ioctl VFIO_DEVICE_GET_IRQ_INFO,
register a event notifier to get the eventfd fd which is sent
to kernel via the ioctl VFIO_DEVICE_SET_IRQS, then we can implement
read operation once kernel sends the signal.
Dong Jia Shi [Wed, 17 May 2017 00:48:08 +0000 (02:48 +0200)]
vfio/ccw: get io region info
vfio-ccw provides an MMIO region for I/O operations. We fetch its
information via ioctls here, then we can use it performing I/O
instructions and retrieving I/O results later on.
Dong Jia Shi [Wed, 17 May 2017 00:48:06 +0000 (02:48 +0200)]
s390x/css: device support for s390-ccw passthrough
In order to support subchannels pass-through, we introduce a s390
subchannel device called "s390-ccw" to hold the real subchannel info.
The s390-ccw devices inherit from the abstract CcwDevice which connect
to the existing virtual-css-bus.
Dong Jia Shi [Wed, 17 May 2017 00:48:05 +0000 (02:48 +0200)]
s390x/css: realize css_create_sch
The S390 virtual css support already has a mechanism to create a
virtual subchannel and provide it to the guest. However, to
pass-through subchannels to a guest, we need to introduce a new
mechanism to create the subchannel according to the real device
information. Thus we reconstruct css_create_virtual_sch to a new
css_create_sch function to handle all these cases and do allocation
and initialization of the subchannel according to the device type
and machine configuration.
The S390 virtual css support already has a mechanism to build a
virtual subchannel information block (schib) and provide virtual
subchannels to the guest. However, to pass-through subchannels to
a guest, we need to introduce a new mechanism to build its schib
according to the real device information. Thus we realize a new css
sch_build_schib function to extract the path_masks, chpids, chpid
type from sysfs. To reuse the existing code, we refactor
css_add_virtual_chpid to css_add_chpid.
We want to support real (i.e. not virtual) channel devices
even for guests that do not support MCSS-E (where guests may
see devices from any channel subsystem image at once). As all
virtio-ccw devices are in css 0xfe (and show up in the default
css 0 for guests not activating MCSS-E), we need an option to
squash both the virtio subchannels and e.g. passed-through
subchannels from their real css (0-3, or 0 for hosts not
activating MCSS-E) into the default css. This will be
exploited in a later patch.
Eric Farman [Wed, 10 May 2017 15:53:59 +0000 (17:53 +0200)]
pc-bios/s390-ccw.img: rebuild image
Contains the following commits:
- pc-bios/s390-ccw: Remove duplicate blk_factor adjustment
- pc-bios/s390-ccw: Move SCSI block factor to outer read
- pc-bios/s390-ccw: Break up virtio-scsi read into multiples
- pc-bios/s390-ccw: Refactor scsi_inquiry function
- pc-bios/s390-ccw: Get list of supported EVPD pages
- pc-bios/s390-ccw: Get Block Limits VPD device data
- pc-bios/s390-ccw: Build a reasonable max_sectors limit
Eric Farman [Wed, 10 May 2017 15:53:58 +0000 (17:53 +0200)]
pc-bios/s390-ccw: Build a reasonable max_sectors limit
Now that we've read all the possible limits that have been defined for
a virtio-scsi controller and the disk we're booting from, it's possible
that we are STILL going to exceed the limits of the host device.
For example, a "-device scsi-generic" device does not support the
Block Limits VPD page.
So, let's fallback to something that seems to work for most boot
configurations if larger values were specified (including if nothing
was explicitly specified, and we took default values).
Eric Farman [Wed, 10 May 2017 15:53:57 +0000 (17:53 +0200)]
pc-bios/s390-ccw: Get Block Limits VPD device data
The "Block Limits" Inquiry VPD page is optional for any SCSI device,
but if it's supported it provides a hint of the maximum I/O transfer
length for this particular device. If this page is supported by the
disk, let's issue that Inquiry and use the minimum of it and the
SCSI controller limit. That will cover this scenario:
Now that we have a limit for a virtio-scsi disk, compare that with the
limit for the virtio-scsi controller when we actually build the I/O.
The minimum of these two limits should be the one we use.
Eric Farman [Wed, 10 May 2017 15:53:56 +0000 (17:53 +0200)]
pc-bios/s390-ccw: Get list of supported VPD pages
The "Supported Pages" Inquiry EVPD page is mandatory for all SCSI devices,
and is used as a gateway for what VPD pages the device actually supports.
Let's issue this Inquiry, and dump that list with the debug facility.
Eric Farman [Wed, 10 May 2017 15:53:55 +0000 (17:53 +0200)]
pc-bios/s390-ccw: Refactor scsi_inquiry function
If we want to issue any of the SCSI Inquiry EVPD pages,
which we do, we could use this function to issue both types
of commands with a little bit of refactoring.
Eric Farman [Wed, 10 May 2017 15:53:54 +0000 (17:53 +0200)]
pc-bios/s390-ccw: Break up virtio-scsi read into multiples
A virtio-scsi request that goes through the host sd driver and exceeds
the maximum transfer size is automatically broken up for us. But the
equivalent request going to the sg driver presumes that any length
requirements have already been honored.
Let's use the max_sectors field on the virtio-scsi controller device,
and break up all requests (both sd and sg) to avoid this problem.
When using virtio-scsi, we multiply the READ(10) data_size by
a block factor twice when building the I/O. This is fine,
since it's only 1 for SCSI disks, but let's clean it up.
Eduardo Habkost [Mon, 8 May 2017 20:57:34 +0000 (17:57 -0300)]
audio: Rename audio_init() to soundhw_init()
To make it consistent with the remaining soundhw.c functions and
avoid confusion with the audio_init() function in audio/audio.c,
rename audio_init() to soundhw_init().
Many users now prefer to use drive_mirror over NBD as an
alternative to the older migrate -b option; drive_mirror is
more complex to setup but gives you more options (e.g. only
migrating some of the disks if some of them are shared).
Allow the large chunk of block migration code to be compiled
out for those who don't use it.
Based on a downstream-patch we've had for a while by Jeff Cody.
Juan Quintela [Wed, 5 Apr 2017 18:45:22 +0000 (20:45 +0200)]
migration: Remove use of old MigrationParams
We have change in the previous patch to use migration capabilities for
it. Notice that we continue using the old command line flags from
migrate command from the time being. Remove the set_params method as
now it is empty.
For savevm, one can't do a:
savevm -b/-i foo
but now one can do:
migrate_set_capability block on
savevm foo
And we can't use block migration. We could disable block capability
unconditionally, but it would not be much better.
Signed-off-by: Juan Quintela <[email protected]> Reviewed-by: Eric Blake <[email protected]>
---
- Maintain shared/enabled dependency (Xu suggestion)
- Now we maintain the dependency on the setter functions
- improve error messages
It turns out that it's legal to create a VM with RAMBlocks that aren't
a multiple of the pagesize in use; e.g. a 1025M main memory using
2M host pages. That breaks postcopy's atomic placement of pages,
so disallow it.
migration: Fix non-multiple of page size migration
Unfortunately it's legal to create a VM with a RAM size that's
not a multiple of the underlying host page or huge page size.
Recently I'd changed things to always send host sized pages,
and that breaks if we have say a 1025MB guest on 2MB hugepages.
Unfortunately we can't just make that illegal since it would break
migration from/to existing oddly configured VMs.
Symptom: qemu-system-x86_64: Illegal RAM offset 40100000
as it transmits the fraction of the hugepage after the end
of the RAMBlock (may also cause a crash on the source
- possibly due to clearing bits after the bitmap)
Reported-by: Yumei Huang <[email protected]>
Red Hat bug: https://bugzilla.redhat.com/show_bug.cgi?id=1449037
Stefan Hajnoczi [Thu, 18 May 2017 09:16:38 +0000 (10:16 +0100)]
Merge remote-tracking branch 'dgilbert/tags/pull-hmp-20170517' into staging
HMP pull
# gpg: Signature made Wed 17 May 2017 07:03:39 PM BST
# gpg: using RSA key 0x0516331EBC5BFDE7
# gpg: Good signature from "Dr. David Alan Gilbert (RH2) <[email protected]>"
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 45F5 C71B 4A0C B7FB 977A 9FA9 0516 331E BC5B FDE7
* dgilbert/tags/pull-hmp-20170517:
ramblock: add new hmp command "info ramblock"
utils: provide size_to_str()
ramblock: add RAMBLOCK_FOREACH()
Stefan Hajnoczi [Thu, 18 May 2017 09:01:00 +0000 (10:01 +0100)]
Merge remote-tracking branch 'mst/tags/for_upstream' into staging
pci, virtio, vhost: fixes
A bunch of fixes that missed the release.
Most notably we are reverting shpc back to enabled by default state
as guests uses that as an indicator that hotplug is supported
(even though it's unused). Unfortunately we can't fix this
on the stable branch since that would break migration.
Signed-off-by: Michael S. Tsirkin <[email protected]>
# gpg: Signature made Wed 17 May 2017 10:42:06 PM BST
# gpg: using RSA key 0x281F0DB8D28D5469
# gpg: Good signature from "Michael S. Tsirkin <[email protected]>"
# gpg: aka "Michael S. Tsirkin <[email protected]>"
# Primary key fingerprint: 0270 606B 6F3C DF3D 0B17 0970 C350 3912 AFBE 8E67
# Subkey fingerprint: 5D09 FD08 71C8 F85B 94CA 8A0D 281F 0DB8 D28D 5469
* mst/tags/for_upstream:
exec: abstract address_space_do_translate()
pci: deassert intx when pci device unrealize
virtio: allow broken device to notify guest
Revert "hw/pci: disable pci-bridge's shpc by default"
acpi-defs: clean up open brace usage
ACPI: don't call acpi_pcihp_device_plug_cb on xen
iommu: Don't crash if machine is not PC_MACHINE
pc: add 2.10 machine type
pc/fwcfg: unbreak migration from qemu-2.5 and qemu-2.6 during firmware boot
libvhost-user: fix crash when rings aren't ready
hw/virtio: fix vhost user fails to startup when MQ
hw/arm/virt: generate 64-bit addressable ACPI objects
hw/acpi-defs: replace leading X with x_ in FADT field names
Peter Xu [Wed, 17 May 2017 08:57:42 +0000 (16:57 +0800)]
exec: abstract address_space_do_translate()
This function is an abstraction helper for address_space_translate() and
address_space_get_iotlb_entry(). It does the lookup of address into
memory region section, then does proper IOMMU translation if necessary.
Refactor the two existing functions to use it.
If a pci device is not reset by VM (by writing into config space)
and unplugged by VM, after that when VM reboots, qemu may assert:
pcibus_reset: Assertion `bus->irq_count[i] == 0' failed
Greg Kurz [Wed, 17 May 2017 08:17:51 +0000 (10:17 +0200)]
virtio: allow broken device to notify guest
According to section 2.1.2 of the virtio-1 specification:
"The device SHOULD set DEVICE_NEEDS_RESET when it enters an error state that
a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET,
the device MUST send a device configuration change notification to the
driver."
Commit "f5ed36635d8f virtio: stop virtqueue processing if device is broken"
introduced a virtio_error() call that just does that:
- internally mark the device as broken
- set the DEVICE_NEEDS_RESET bit in the status
- send a configuration change notification
Unfortunately, virtio_notify_vector(), called by virtio_notify_config(),
returns right away when the device is marked as broken and the notification
isn't sent in this case.
The spec doesn't say whether a broken device can send notifications
in other situations or not. But since the driver isn't supposed to do
anything but to reset the device, it makes sense to keep the check in
virtio_notify_config().
Marking the device as broken AFTER the configuration change notification was
sent is enough to fix the issue.
Disabling the shpc controller has an undesired side effect.
The PCI bridge remains with no attached devices at boot time,
and the guest operating systems do not allocate any resources
for it, leaving the bridge unusable. Note that the behaviour
is dictated by the pci bridge specification.
Revert the commit and leave the shpc controller even if is not
actually used by any architecture. Slot 0 remains unusable at boot time.
Ramblock is something hidden internally in QEMU implementation, and this
command should only be used by mostly QEMU developers on RAM stuff. It
is not a command suitable for QMP interface. So only HMP interface is
provided for it.
Stefan Hajnoczi [Wed, 17 May 2017 15:34:32 +0000 (16:34 +0100)]
Merge remote-tracking branch 'ehabkost/tags/x86-and-machine-pull-request' into staging
x86 and machine queue, 2017-05-17
# gpg: Signature made Wed 17 May 2017 02:37:54 PM BST
# gpg: using RSA key 0x2807936F984DC5A6
# gpg: Good signature from "Eduardo Habkost <[email protected]>"
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF D1AA 2807 936F 984D C5A6
* ehabkost/tags/x86-and-machine-pull-request: (22 commits)
tests: Add [+-]feature and feature=on|off test cases
s390-pcibus: No need to set user_creatable=false explicitly
xen-sysdev: Remove user_creatable flag
virtio-mmio: Remove user_creatable flag
sysbus-ohci: Remove user_creatable flag
hpet: Remove user_creatable flag
generic-sdhci: Remove user_creatable flag
esp: Remove user_creatable flag
fw_cfg: Remove user_creatable flag
unimplemented-device: Remove user_creatable flag
isabus-bridge: Remove user_creatable flag
allwinner-ahci: Remove user_creatable flag
sysbus-ahci: Remove user_creatable flag
kvmvapic: Remove user_creatable flag
ioapic: Remove user_creatable flag
kvmclock: Remove user_creatable flag
pflash_cfi01: Remove user_creatable flag
fdc: Remove user_creatable flag from sysbus-fdc & SUNW,fdtwo
iommu: Remove FIXME comment about user_creatable=true
xen-backend: Remove FIXME comment about user_creatable flag
...
Eduardo Habkost [Mon, 8 May 2017 18:32:05 +0000 (15:32 -0300)]
tests: Add [+-]feature and feature=on|off test cases
Add test code to ensure features are enabled/disabled correctly in the
command-line. The test case use the "feature-words" and
"filtered-features" properties to check if the features were
enabled/disabled correctly.
Eduardo Habkost [Wed, 3 May 2017 20:36:04 +0000 (17:36 -0300)]
s390-pcibus: No need to set user_creatable=false explicitly
TYPE_S390_PCI_HOST_BRIDGE is a subclass of TYPE_PCI_HOST_BRIDGE,
which is a subclass of TYPE_SYS_BUS_DEVICE. TYPE_SYS_BUS_DEVICE
already sets user_creatable=false, so we don't require an
explicit user_creatable=false assignment in
s390_pcihost_class_init().
Eduardo Habkost [Wed, 3 May 2017 20:36:03 +0000 (17:36 -0300)]
xen-sysdev: Remove user_creatable flag
TYPE_XENSYSDEV is only used internally by xen_be_init(), and is
not supposed to be plugged/unplugged dynamically. Remove the
user_creatable flag from the device class.
Eduardo Habkost [Wed, 3 May 2017 20:36:02 +0000 (17:36 -0300)]
virtio-mmio: Remove user_creatable flag
virtio-mmio needs to be wired and mapped by other device or board
code, and won't work with -device. Remove the user_creatable flag
from the device class.
Eduardo Habkost [Wed, 3 May 2017 20:35:58 +0000 (17:35 -0300)]
esp: Remove user_creatable flag
esp devices aren't going to work with -device, as they need IRQs
to be connected and mmio to be mapped (this is done by
esp_init()). Remove the user_creatable flag from the device
class.
Eduardo Habkost [Wed, 3 May 2017 20:35:57 +0000 (17:35 -0300)]
fw_cfg: Remove user_creatable flag
fw_cfg won't work with -device, as:
* fw_cfg_init1() won't get called for the device;
* The device won't appear at /machine/fw_cfg, and won't work with
the -fw_cfg command-line option.
Remove the user_creatable flag from the device class.
Eduardo Habkost [Wed, 3 May 2017 20:35:56 +0000 (17:35 -0300)]
unimplemented-device: Remove user_creatable flag
unimplemented-device needs to be created and mapped using
create_unimplemented_device() (or equivalent code), and won't
work with -device. Remove the user_creatable flag from the device
class.
Eduardo Habkost [Wed, 3 May 2017 20:35:55 +0000 (17:35 -0300)]
isabus-bridge: Remove user_creatable flag
isabus-bridge needs to be created by isa_bus_new(), and won't
work with -device, as it won't create the TYPE_ISA_BUS bus
itself. Remove the user_creatable flag from the device class.
Eduardo Habkost [Wed, 3 May 2017 20:35:54 +0000 (17:35 -0300)]
allwinner-ahci: Remove user_creatable flag
allwinner-ahci needs its IRQ to be connected and mmio to be
mapped (this is done by the alwinner-a10 device realize method),
and won't work with -device. Remove the user_creatable flag from
the device class.
Eduardo Habkost [Wed, 3 May 2017 20:35:53 +0000 (17:35 -0300)]
sysbus-ahci: Remove user_creatable flag
The sysbus-ahci devices are supposed to be created and wired by
code from other devices, like calxeda_init() and
xlnx_zynqmp_realize(), and won't work with -device. Remove the
user_creatable flag from the device class.
Eduardo Habkost [Wed, 3 May 2017 20:35:51 +0000 (17:35 -0300)]
ioapic: Remove user_creatable flag
An ioapic device is already created by the q35 initialization
code, and using "-device ioapic" or "-device kvm-ioapic" will
always fail with "Only 1 ioapics allowed". Remove the
user_creatable flag from the ioapic device classes.
Eduardo Habkost [Wed, 3 May 2017 20:35:50 +0000 (17:35 -0300)]
kvmclock: Remove user_creatable flag
kvmclock should be used by guests only when the appropriate CPUID
feature flags are set on the VCPU, and it is automatically
created by kvmclock_create() when those feature flags are set.
This means creating a kvmclock device using -device is useless.
Remove user_creatable from its device class.
Eduardo Habkost [Wed, 3 May 2017 20:35:49 +0000 (17:35 -0300)]
pflash_cfi01: Remove user_creatable flag
TYPE_CFI_PFLASH01 devices need to be mapped by
pflash_cfi01_register() (or equivalent) and can't be used with
-device. Remove user_creatable from the device class.
Eduardo Habkost [Wed, 3 May 2017 20:35:48 +0000 (17:35 -0300)]
fdc: Remove user_creatable flag from sysbus-fdc & SUNW,fdtwo
sysbus-fdc and SUNW,fdtwo devices need IRQs to be wired and mmio
to be mapped, and can't be used with -device. Unset
user_creatable on their device classes.
Eduardo Habkost [Wed, 3 May 2017 20:35:46 +0000 (17:35 -0300)]
xen-backend: Remove FIXME comment about user_creatable flag
xen-backend can be plugged/unplugged dynamically when using the
Xen accelerator, so keep the user_creatable flag on the device
class and remove the FIXME comment.
Eduardo Habkost [Wed, 3 May 2017 20:35:45 +0000 (17:35 -0300)]
sysbus: Set user_creatable=false by default on TYPE_SYS_BUS_DEVICE
commit 33cd52b5d7b9adfd009e95f07e6c64dd88ae2a31 unset
cannot_instantiate_with_device_add_yet in TYPE_SYSBUS, making all
sysbus devices appear on "-device help" and lack the "no-user"
flag in "info qdm".
To fix this, we can set user_creatable=false by default on
TYPE_SYS_BUS_DEVICE, but this requires setting
user_creatable=true explicitly on the sysbus devices that
actually work with -device.
Fortunately today we have just a few has_dynamic_sysbus=1
machines: virt, pc-q35-*, ppce500, and spapr.
virt, ppce500, and spapr have extra checks to ensure just a few
device types can be instantiated:
* virt supports only TYPE_VFIO_CALXEDA_XGMAC, TYPE_VFIO_AMD_XGBE.
* ppce500 supports only TYPE_ETSEC_COMMON.
* spapr supports only TYPE_SPAPR_PCI_HOST_BRIDGE.
This patch sets user_creatable=true explicitly on those 4 device
classes.
Now, the more complex cases:
pc-q35-*: q35 has no sysbus device whitelist yet (which is a
separate bug). We are in the process of fixing it and building a
sysbus whitelist on q35, but in the meantime we can fix the
"-device help" and "info qdm" bugs mentioned above. Also, despite
not being strictly necessary for fixing the q35 bug, reducing the
list of user_creatable=true devices will help us be more
confident when building the q35 whitelist.
xen: We also have a hack at xen_set_dynamic_sysbus(), that sets
has_dynamic_sysbus=true at runtime when using the Xen
accelerator. This hack is only used to allow xen-backend devices
to be dynamically plugged/unplugged.
This means today we can use -device with the following 22 device
types, that are the ones compiled into the qemu-system-x86_64 and
qemu-system-i386 binaries:
This patch adds user_creatable=true explicitly to those devices,
temporarily, just to keep 100% compatibility with existing
behavior of q35. Subsequent patches will remove
user_creatable=true from the devices that are really not meant to
user-creatable on any machine, and remove the FIXME comment from
the ones that are really supposed to be user-creatable. This is
being done in separate patches because we still don't have an
obvious list of devices that will be whitelisted by q35, and I
would like to get each device reviewed individually.
Eduardo Habkost [Wed, 3 May 2017 20:35:44 +0000 (17:35 -0300)]
qdev: Replace cannot_instantiate_with_device_add_yet with !user_creatable
cannot_instantiate_with_device_add_yet was introduced by commit efec3dd631d94160288392721a5f9c39e50fb2bc to replace no_user. It was
supposed to be a temporary measure.
When it was introduced, we had 54
cannot_instantiate_with_device_add_yet=true lines in the code.
Today (3 years later) this number has not shrunk: we now have
57 cannot_instantiate_with_device_add_yet=true lines. I think it
is safe to say it is not a temporary measure, and we won't see
the flag go away soon.
Instead of a long field name that misleads people to believe it
is temporary, replace it a shorter and less misleading field:
user_creatable.
Except for code comments, changes were generated using the
following Coccinelle patch:
Stefan Hajnoczi [Wed, 17 May 2017 13:03:32 +0000 (14:03 +0100)]
Merge remote-tracking branch 'sstabellini/tags/xen-20170516-tag' into staging
Xen 2017/05/16
# gpg: Signature made Tue 16 May 2017 08:18:32 PM BST
# gpg: using RSA key 0x894F8F4870E1AE90
# gpg: Good signature from "Stefano Stabellini <[email protected]>"
# gpg: aka "Stefano Stabellini <[email protected]>"
# Primary key fingerprint: D04E 33AB A51F 67BA 07D3 0AEA 894F 8F48 70E1 AE90
* sstabellini/tags/xen-20170516-tag:
xen: call qemu_set_cloexec instead of fcntl
xen/9pfs: fix two resource leaks on error paths, discovered by Coverity
configure: Remove -lxencall for Xen detection
xen/mapcache: store dma information in revmapcache entries for debugging
* jtc/tags/block-pull-request:
curl: do not do aio_poll when waiting for a free CURLState
curl: convert readv to coroutines
curl: convert CURLAIOCB to byte values
curl: split curl_find_state/curl_init_state
curl: avoid recursive locking of BDRVCURLState mutex
curl: never invoke callbacks with s->mutex held
curl: strengthen assertion in curl_clean_state
block: curl: Allow passing cookies via QCryptoSecret
Juan Quintela [Mon, 17 Apr 2017 12:57:54 +0000 (14:57 +0200)]
migration: Move check_migratable() into qdev.c
The function is only used once, and nothing else in migration knows
about objects. Create the function vmstate_device_is_migratable() in
savem.c that really do the bit that is related with migration.
On do_compress_ram_page() we use a different QEMUFile than the
migration one. We need to pass it there. The failure can be seen as:
(qemu) qemu-system-x86_64: Unknown combination of migration flags: 0
qemu-system-x86_64: error while loading state section id 3(ram)
qemu-system-x86_64: load of migration failed: Invalid argument
Use the common utility function, which contains checks on return values
and first calls F_GETFD as recommended by POSIX.1-2001, instead of
manually calling fcntl.
Anthony PERARD [Thu, 11 May 2017 11:35:42 +0000 (12:35 +0100)]
configure: Remove -lxencall for Xen detection
QEMU does not depends on libxencall, it was added because it was a
missing link dependency of libxendevicemodel, but now the later should
be built properly.
xen/mapcache: store dma information in revmapcache entries for debugging
The Xen mapcache is able to create long term mappings, they are called
"locked" mappings. The third parameter of the xen_map_cache call
specifies if a mapping is a "locked" mapping.
>From the QEMU point of view there are two kinds of long term mappings:
[a] device memory mappings, such as option roms and video memory
[b] dma mappings, created by dma_memory_map & friends
After certain operations, ballooning a VM in particular, Xen asks QEMU
kindly to destroy all mappings. However, certainly [a] mappings are
present and cannot be removed. That's not a problem as they are not
affected by balloonning. The *real* problem is that if there are any
mappings of type [b], any outstanding dma operations could fail. This is
a known shortcoming. In other words, when Xen asks QEMU to destroy all
mappings, it is an error if any [b] mappings exist.
However today we have no way of distinguishing [a] from [b]. Because of
that, we cannot even print a decent warning.
This patch introduces a new "dma" bool field to MapCacheRev entires, to
remember if a given mapping is for dma or is a long term device memory
mapping. When xen_invalidate_map_cache is called, we print a warning if
any [b] mappings exist. We ignore [a] mappings.
Mappings created by qemu_map_ram_ptr are assumed to be [a], while
mappings created by address_space_map->qemu_ram_ptr_length are assumed
to be [b].
The goal of the patch is to make debugging and system understanding
easier.
Paolo Bonzini [Mon, 15 May 2017 10:00:59 +0000 (12:00 +0200)]
curl: do not do aio_poll when waiting for a free CURLState
Instead, put the CURLAIOCB on a wait list and yield; curl_clean_state will
wake the corresponding coroutine.
Because of CURL's callback-based structure, we cannot easily convert
everything to CoMutex/CoQueue; keeping the QemuMutex is simpler. However,
CoQueue is a simple wrapper around a linked list, so we can easily
use QSIMPLEQ and open-code a CoQueue, protected by the BDRVCURLState
QemuMutex instead of a CoMutex.
Paolo Bonzini [Mon, 15 May 2017 10:00:58 +0000 (12:00 +0200)]
curl: convert readv to coroutines
This is pretty simple. The bottom half goes away because, unlike
bdrv_aio_readv, coroutine-based read can return immediately without
yielding. However, for simplicity I kept the former bottom half
handler in a separate function.
Paolo Bonzini [Mon, 15 May 2017 10:00:57 +0000 (12:00 +0200)]
curl: convert CURLAIOCB to byte values
This is in preparation for the conversion from bdrv_aio_readv to
bdrv_co_preadv, and it also requires changing some of the size_t values
to uint64_t. This was broken before for disks > 2TB, but now it would
break at 4GB.
Paolo Bonzini [Mon, 15 May 2017 10:00:56 +0000 (12:00 +0200)]
curl: split curl_find_state/curl_init_state
If curl_easy_init fails, a CURLState is left with s->in_use = 1. Split
curl_init_state in two, so that we can distinguish the two failures and
call curl_clean_state if needed.
While at it, simplify curl_find_state, removing a dummy loop. The
aio_poll loop is moved to the sole caller that needs it.
Paolo Bonzini [Mon, 15 May 2017 10:00:55 +0000 (12:00 +0200)]
curl: avoid recursive locking of BDRVCURLState mutex
The curl driver has a ugly hack where, if it cannot find an empty CURLState,
it just uses aio_poll to wait for one to be empty. This is probably
buggy when used together with dataplane, and the simplest way to fix it
is to use coroutines instead.
A more immediate effect of the bug however is that it can cause a
recursive call to curl_readv_bh_cb and recursively taking the
BDRVCURLState mutex. This causes a deadlock.
The fix is to unlock the mutex around aio_poll, but for cleanliness we
should also take the mutex around all calls to curl_init_state, even if
reaching the unlock/lock pair is impossible. The same is true for
curl_clean_state.
projects / qemu.git / log