| Commit message (Collapse) | Author | Files | Lines |
|---|
|
Prove that blockdev-mirror can now result in sparse raw destination
files, regardless of whether the source is raw or qcow2. By making
this a separate test, it was possible to test effects of individual
patches for the various pieces that all have to work together for a
sparse mirror to be successful.
Note that ./check -file produces different job lengths than ./check
-qcow2 (the test uses a filter to normalize); that's because when
deciding how much of the image to be mirrored, the code looks at how
much of the source image was allocated (for qcow2, this is only the
written clusters; for raw, it is the entire file). But the important
part is that the destination file ends up smaller than 3M, rather than
the 20M it used to be before this patch series.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-ID: <20250509204341.3553601-28-eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
|
|
Move the definition from iotests/250 to common.rc. This is used to
detect real disk usage of sparse files. In particular, we want to use
it for checking subclusters-based discards.
Signed-off-by: Andrey Drobyshev <andrey.drobyshev@virtuozzo.com>
Reviewed-by: Alexander Ivanov <alexander.ivanov@virtuozzo.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-ID: <20240913163942.423050-6-andrey.drobyshev@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20250509204341.3553601-27-eblake@redhat.com>
|
|
When mirroring, the goal is to ensure that the destination reads the
same as the source; this goal is met whether the destination is sparse
or fully-allocated (except when explicitly punching holes, then merely
reading zero is not enough to know if it is sparse, so we still want
to punch the hole). Avoiding a redundant write to zero (whether in
the background because the zero cluster was marked in the dirty
bitmap, or in the foreground because the guest is writing zeroes) when
the destination already reads as zero makes mirroring faster, and
avoids allocating the destination merely because the source reports as
allocated.
The effect is especially pronounced when the source is a raw file.
That's because when the source is a qcow2 file, the dirty bitmap only
visits the portions of the source that are allocated, which tend to be
non-zero. But when the source is a raw file,
bdrv_co_is_allocated_above() reports the entire file as allocated so
mirror_dirty_init sets the entire dirty bitmap, and it is only later
during mirror_iteration that we change to consulting the more precise
bdrv_co_block_status_above() to learn where the source reads as zero.
Remember that since a mirror operation can write a cluster more than
once (every time the guest changes the source, the destination is also
changed to keep up), and the guest can change whether a given cluster
reads as zero, is discarded, or has non-zero data over the course of
the mirror operation, we can't take the shortcut of relying on
s->target_is_zero (which is static for the life of the job) in
mirror_co_zero() to see if the destination is already zero, because
that information may be stale. Any solution we use must be dynamic in
the face of the guest writing or discarding a cluster while the mirror
has been ongoing.
We could just teach mirror_co_zero() to do a block_status() probe of
the destination, and skip the zeroes if the destination already reads
as zero, but we know from past experience that extra block_status()
calls are not always cheap (tmpfs, anyone?), especially when they are
random access rather than linear. Use of block_status() of the source
by the background task in a linear fashion is not our bottleneck (it's
a background task, after all); but since mirroring can be done while
the source is actively being changed, we don't want a slow
block_status() of the destination to occur on the hot path of the
guest trying to do random-access writes to the source.
So this patch takes a slightly different approach: any time we have to
track dirty clusters, we can also track which clusters are known to
read as zero. For sync=TOP or when we are punching holes from
"detect-zeroes":"unmap", the zero bitmap starts out empty, but
prevents a second write zero to a cluster that was already zero by an
earlier pass; for sync=FULL when we are not punching holes, the zero
bitmap starts out full if the destination reads as zero during
initialization. Either way, I/O to the destination can now avoid
redundant write zero to a cluster that already reads as zero, all
without having to do a block_status() per write on the destination.
With this patch, if I create a raw sparse destination file, connect it
with QMP 'blockdev-add' while leaving it at the default "discard":
"ignore", then run QMP 'blockdev-mirror' with "sync": "full", the
destination remains sparse rather than fully allocated. Meanwhile, a
destination image that is already fully allocated remains so unless it
was opened with "detect-zeroes": "unmap". And any time writing zeroes
is skipped, the job counters are not incremented.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-ID: <20250509204341.3553601-26-eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
|
|
When doing a sync=full mirroring, we can skip pre-zeroing the
destination if it already reads as zeroes and we are not also trying
to punch holes due to detect-zeroes. With this patch, there are fewer
scenarios that have to pass in an explicit target-is-zero, while still
resulting in a sparse destination remaining sparse.
A later patch will then further improve things to skip writing to the
destination for parts of the image where the source is zero; but even
with just this patch, it is possible to see a difference for any
source that does not report itself as fully allocated, coupled with a
destination BDS that can quickly report that it already reads as zero.
(For a source that reports as fully allocated, such as a file, the
rest of mirror_dirty_init() still sets the entire dirty bitmap to
true, so even though we avoided the pre-zeroing, we are not yet
avoiding all redundant I/O).
Iotest 194 detects the difference made by this patch: for a file
source (where block status reports the entire image as allocated, and
therefore we end up writing zeroes everywhere in the destination
anyways), the job length remains the same. But for a qcow2 source and
a destination that reads as all zeroes, the dirty bitmap changes to
just tracking the allocated portions of the source, which results in
faster completion and smaller job statistics. For the test to pass
with both ./check -file and -qcow2, a new python filter is needed to
mask out the now-varying job amounts (this matches the shell filters
_filter_block_job_{offset,len} in common.filter). A later test will
also be added which further validates expected sparseness, so it does
not matter that 194 is no longer explicitly looking at how many bytes
were copied.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-ID: <20250509204341.3553601-25-eblake@redhat.com>
Reviewed-by: Sunny Zhu <sunnyzhyy@qq.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
|
|
The two callers to a mirror job (drive-mirror and blockdev-mirror) set
zero_target precisely when sync mode == FULL, with the one exception
that drive-mirror skips zeroing the target if it was newly created and
reads as zero. But given the previous patch, that exception is
equally captured by target_is_zero.
Meanwhile, there is another slight wrinkle, fortunately caught by
iotest 185: if the caller uses "sync":"top" but the source has no
backing file, the code in blockdev.c was changing sync to be FULL, but
only after it had set zero_target=false. In mirror.c, prior to recent
patches, this didn't matter: the only places that inspected sync were
setting is_none_mode (both TOP and FULL had set that to false), and
mirror_start() setting base = mode == MIRROR_SYNC_MODE_TOP ?
bdrv_backing_chain_next(bs) : NULL. But now that we are passing sync
around, the slammed sync mode would result in a new pre-zeroing pass
even when the user had passed "sync":"top" in an effort to skip
pre-zeroing. Fortunately, the assignment of base when bs has no
backing chain still works out to NULL if we don't slam things. So
with the forced change of sync ripped out of blockdev.c, the sync mode
is passed through the full callstack unmolested, and we can now
reliably reconstruct the same settings as what used to be passed in by
zero_target=false, without the redundant parameter.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-ID: <20250509204341.3553601-24-eblake@redhat.com>
Reviewed-by: Sunny Zhu <sunnyzhyy@qq.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
[eblake: Fix regression in iotest 185]
Signed-off-by: Eric Blake <eblake@redhat.com>
|
|
QEMU has an optimization for a just-created drive-mirror destination
that is not possible for blockdev-mirror (which can't create the
destination) - any time we know the destination starts life as all
zeroes, we can skip a pre-zeroing pass on the destination. Recent
patches have added an improved heuristic for detecting if a file
contains all zeroes, and we plan to use that heuristic in upcoming
patches. But since a heuristic cannot quickly detect all scenarios,
and there may be cases where the caller is aware of information that
QEMU cannot learn quickly, it makes sense to have a way to tell QEMU
to assume facts about the destination that can make the mirror
operation faster. Given our existing example of "qemu-img convert
--target-is-zero", it is time to expose this override in QMP for
blockdev-mirror as well.
This patch results in some slight redundancy between the older
s->zero_target (set any time mode==FULL and the destination image was
not just created - ie. clear if drive-mirror is asking to skip the
pre-zero pass) and the newly-introduced s->target_is_zero (in addition
to the QMP override, it is set when drive-mirror creates the
destination image); this will be cleaned up in the next patch.
There is also a subtlety that we must consider. When drive-mirror is
passing target_is_zero on behalf of a just-created image, we know the
image is sparse (skipping the pre-zeroing keeps it that way), so it
doesn't matter whether the destination also has "discard":"unmap" and
"detect-zeroes":"unmap". But now that we are letting the user set the
knob for target-is-zero, if the user passes a pre-existing file that
is fully allocated, it is fine to leave the file fully allocated under
"detect-zeroes":"on", but if the file is open with
"detect-zeroes":"unmap", we should really be trying harder to punch
holes in the destination for every region of zeroes copied from the
source. The easiest way to do this is to still run the pre-zeroing
pass (turning the entire destination file sparse before populating
just the allocated portions of the source), even though that currently
results in double I/O to the portions of the file that are allocated.
A later patch will add further optimizations to reduce redundant
zeroing I/O during the mirror operation.
Since "target-is-zero":true is designed for optimizations, it is okay
to silently ignore the parameter rather than erroring if the user ever
sets the parameter in a scenario where the mirror job can't exploit it
(for example, when doing "sync":"top" instead of "sync":"full", we
can't pre-zero, so setting the parameter won't make a speed
difference).
Signed-off-by: Eric Blake <eblake@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Message-ID: <20250509204341.3553601-23-eblake@redhat.com>
Reviewed-by: Sunny Zhu <sunnyzhyy@qq.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
|
|
Out of the five possible values for MirrorSyncMode, INCREMENTAL and
BITMAP are already rejected up front in mirror_start, leaving NONE,
TOP, and FULL as the remaining values that the code was collapsing
into a single bool is_none_mode. Furthermore, mirror_dirty_init() is
only reachable for modes TOP and FULL, as further guided by
s->zero_target. However, upcoming patches want to further optimize
the pre-zeroing pass of a sync=full mirror in mirror_dirty_init(),
while avoiding that pass on a sync=top action. Instead of throwing
away context by collapsing these two values into
s->is_none_mode=false, it is better to pass s->sync_mode throughout
the entire operation. For active commit, the desired semantics match
sync mode TOP.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-ID: <20250509204341.3553601-22-eblake@redhat.com>
Reviewed-by: Sunny Zhu <sunnyzhyy@qq.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
|
|
Commit 5791ba52 (v9.2) pre-initialized ret in mirror_dirty_init to
silence a false positive compiler warning, even though in all code
paths where ret is used, it was guaranteed to be reassigned
beforehand. But since the function returns -errno, and -1 is not
always the right errno, it's better to initialize to -EIO.
An upcoming patch wants to track two bitmaps in
do_sync_target_write(); this will be easier if the current variables
related to the dirty bitmap are renamed.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20250509204341.3553601-21-eblake@redhat.com>
|
|
Mirroring a completely sparse image to a sparse destination should be
practically instantaneous. It isn't yet, but the test will be more
realistic if it has some non-zero to mirror as well as the holes.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20250509204341.3553601-20-eblake@redhat.com>
|
|
There are some optimizations that require knowing if an image starts
out as reading all zeroes, such as making blockdev-mirror faster by
skipping the copying of source zeroes to the destination. The
existing bdrv_co_is_zero_fast() is a good building block for answering
this question, but it tends to give an answer of 0 for a file we just
created via QMP 'blockdev-create' or similar (such as 'qemu-img create
-f raw'). Why? Because file-posix.c insists on allocating a tiny
header to any file rather than leaving it 100% sparse, due to some
filesystems that are unable to answer alignment probes on a hole. But
teaching file-posix.c to read the tiny header doesn't scale - the
problem of a small header is also visible when libvirt sets up an NBD
client to a just-created file on a migration destination host.
So, we need a wrapper function that handles a bit more complexity in a
common manner for all block devices - when the BDS is mostly a hole,
but has a small non-hole header, it is still worth the time to read
that header and check if it reads as all zeroes before giving up and
returning a pessimistic answer.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20250509204341.3553601-19-eblake@redhat.com>
|
|
Some BDS drivers have a cap on how much block status they can supply
in one query (for example, NBD talking to an older server cannot
inspect more than 4G per query; and qcow2 tends to cap its answers
rather than cross a cluster boundary of an L1 table). Although the
existing callers of bdrv_co_is_zero_fast are not passing in that large
of a 'bytes' parameter, an upcoming caller wants to query the entire
image at once, and will thus benefit from being able to treat adjacent
zero regions in a coalesced manner, rather than claiming the region is
non-zero merely because pnum was truncated and didn't match the
incoming bytes.
While refactoring this into a loop, note that there is no need to
assign pnum prior to calling bdrv_co_common_block_status_above() (it
is guaranteed to be assigned deeper in the callstack).
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20250509204341.3553601-18-eblake@redhat.com>
|
|
Although the previous patch to change 'bool want_zero' into a bitmask
made no semantic change, it is now time to differentiate. When the
caller specifically wants to know what parts of the file read as zero,
we need to use lseek and actually reporting holes, rather than
short-circuiting and advertising full allocation.
This change will be utilized in later patches to let mirroring
optimize for the case when the destination already reads as zeroes.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20250509204341.3553601-17-eblake@redhat.com>
|
|
This patch is purely mechanical, changing bool want_zero into an
unsigned int for bitwise-or of flags. As of this patch, all
implementations are unchanged (the old want_zero==true is now
mode==BDRV_WANT_PRECISE which is a superset of BDRV_WANT_ZERO); but
the callers in io.c that used to pass want_zero==false are now
prepared for future driver changes that can now distinguish bewteen
BDRV_WANT_ZERO vs. BDRV_WANT_ALLOCATED. The next patch will actually
change the file-posix driver along those lines, now that we have
more-specific hints.
As for the background why this patch is useful: right now, the
file-posix driver recognizes that if allocation is being queried, the
entire image can be reported as allocated (there is no backing file to
refer to) - but this throws away information on whether the entire
image reads as zero (trivially true if lseek(SEEK_HOLE) at offset 0
returns -ENXIO, a bit more complicated to prove if the raw file was
created with 'qemu-img create' since we intentionally allocate a small
chunk of all-zero data to help with alignment probing). Later patches
will add a generic algorithm for seeing if an entire file reads as
zeroes.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-ID: <20250509204341.3553601-16-eblake@redhat.com>
|
|
Add migration support for AMD IOMMU model by saving necessary AMDVIState
parameters for MMIO registers, device table, command buffer, and event
buffers.
Also change devtab_len type from size_t to uint64_t to avoid 32-bit build
issue.
Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Message-Id: <20250504170405.12623-3-suravee.suthikulpanit@amd.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
control over the PCI device creation
Current amd-iommu model internally creates an AMDVI-PCI device. Here is
a snippet from info qtree:
bus: main-system-bus
type System
dev: amd-iommu, id ""
xtsup = false
pci-id = ""
intremap = "on"
device-iotlb = false
pt = true
...
dev: q35-pcihost, id ""
MCFG = -1 (0xffffffffffffffff)
pci-hole64-size = 34359738368 (32 GiB)
below-4g-mem-size = 134217728 (128 MiB)
above-4g-mem-size = 0 (0 B)
smm-ranges = true
x-pci-hole64-fix = true
x-config-reg-migration-enabled = true
bypass-iommu = false
bus: pcie.0
type PCIE
dev: AMDVI-PCI, id ""
addr = 01.0
romfile = ""
romsize = 4294967295 (0xffffffff)
rombar = -1 (0xffffffffffffffff)
multifunction = false
x-pcie-lnksta-dllla = true
x-pcie-extcap-init = true
failover_pair_id = ""
acpi-index = 0 (0x0)
x-pcie-err-unc-mask = true
x-pcie-ari-nextfn-1 = false
x-max-bounce-buffer-size = 4096 (4 KiB)
x-pcie-ext-tag = true
busnr = 0 (0x0)
class Class 0806, addr 00:01.0, pci id 1022:0000 (sub 1af4:1100)
...
This prohibits users from specifying the PCI topology for the amd-iommu device,
which becomes a problem when trying to support VM migration since it does not
guarantee the same enumeration of AMD IOMMU device.
Therefore, allow the 'AMDVI-PCI' device to optionally be pre-created and
associated with a 'amd-iommu' device via a new 'pci-id' parameter on the
latter.
For example:
-device AMDVI-PCI,id=iommupci0,bus=pcie.0,addr=0x05 \
-device amd-iommu,intremap=on,pt=on,xtsup=on,pci-id=iommupci0 \
For backward-compatibility, internally create the AMDVI-PCI device if not
specified on the CLI.
Co-developed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
Message-Id: <20250504170405.12623-2-suravee.suthikulpanit@amd.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
vtd_find_add_as can be called by multiple threads which leads to a race
condition. Taking the IOMMU lock ensures we avoid such a race.
Moreover we also need to take the bql to avoid an assert to fail in
memory_region_add_subregion_overlap when actually allocating a new
address space.
Signed-off-by: Clement Mathieu--Drif <clement.mathieu--drif@eviden.com>
Message-Id: <20250430124750.240412-3-clement.mathieu--drif@eviden.com>
Reviewed-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
It is possible nowadays to compile QEMU without pixman support - in that
case the screendump command is not available and the related tests fail.
Thus skip these tests if the screendump command could not be executed.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20250325081713.283490-2-thuth@redhat.com>
|
|
The s390x tuxrun test lacks the call to self.set_machine(), so this
test is currently failing in case the 's390-ccw-virtio' machine has
not been compiled into the binary. Add the check now to fix it.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20250424090640.664217-1-thuth@redhat.com>
|
|
console.h brings a dependency on the <epoxy/opengl.h> and the pixman
header file (if available), so we should avoid to include this file
if it is not really necessary. console.h does not seem to be necessary
for the xlnx_dpdma code, so drop the include here.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20250508144120.163009-2-thuth@redhat.com>
|
|
vtd_switch_address_space needs to take the BQL if not already held.
Use BQL_LOCK_GUARD to make the iommu implementation more consistent.
Signed-off-by: Clement Mathieu--Drif <clement.mathieu--drif@eviden.com>
Message-Id: <20250430124750.240412-2-clement.mathieu--drif@eviden.com>
Reviewed-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
Move virtio_reset() to a later part of the file to remove the forward
declaration of virtio_set_features_nocheck() and to prepare the
situation that we need more operations to perform during reset.
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250421-reset-v2-2-e4c1ead88ea1@daynix.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
virtio-net expects set_features() will be called when the feature set
used by the guest changes to update the number of virtqueues but it is
not called during reset, which will clear all features, leaving the
queues added for VIRTIO_NET_F_MQ or VIRTIO_NET_F_RSS. Not only these
extra queues are visible to the guest, they will cause segmentation
fault during migration.
Call set_features() during reset to remove those queues for virtio-net
as we call set_status(). It will also prevent similar bugs for
virtio-net and other devices in the future.
Fixes: f9d6dbf0bf6e ("virtio-net: remove virtio queues if the guest doesn't support multiqueue")
Buglink: https://issues.redhat.com/browse/RHEL-73842
Cc: qemu-stable@nongnu.org
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250421-reset-v2-1-e4c1ead88ea1@daynix.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
So far there isn't way to test host kernel vhost-scsi event queue path,
because VIRTIO_SCSI_F_HOTPLUG isn't supported by QEMU.
virtio-scsi.c and vhost-user-scsi.c already support VIRTIO_SCSI_F_HOTPLUG
as property "hotplug".
Add support to vhost-scsi.c to help evaluate and test event queue.
To test the feature:
1. Create vhost-scsi target with targetcli.
targetcli /backstores/fileio create name=storage file_or_dev=disk01.raw
targetcli /vhost create naa.1123451234512345
targetcli /vhost/naa.1123451234512345/tpg1/luns create /backstores/fileio/storage
2. Create QEMU instance with vhost-scsi.
-device vhost-scsi-pci,wwpn=naa.1123451234512345,hotplug=true
3. Once guest bootup, hotplug a new LUN from host.
targetcli /backstores/fileio create name=storage02 file_or_dev=disk02.raw
targetcli /vhost/naa.1123451234512345/tpg1/luns create /backstores/fileio/storage02
Signed-off-by: Dongli Zhang <dongli.zhang@oracle.com>
Message-Id: <20250203005215.1502-1-dongli.zhang@oracle.com>
Acked-by: Stefano Garzarella <sgarzare@redhat.com>
|
|
Live migration should be terminated if the vhost-user backend crashes
before the migration completes.
Specifically, since the vhost device will be stopped when VM is stopped
before the end of the live migration, in current implementation if the
backend crashes, vhost-user device set_status() won't return failure,
live migration won't perceive the disconnection between QEMU and the
backend.
When the VM is migrated to the destination, the inflight IO will be
resubmitted, and if the IO was completed out of order before, it will
cause IO error.
To fix this issue:
1. Add the return value to set_status() for VirtioDeviceClass.
a. For the vhost-user device, return failure when the backend crashes.
b. For other virtio devices, always return 0.
2. Return failure if vhost_dev_stop() failed for vhost-user device.
If QEMU loses connection with the vhost-user backend, virtio set_status()
can return failure to the upper layer, migration_completion() can handle
the error, terminate the live migration, and restore the VM, so that
inflight IO can be completed normally.
Signed-off-by: Haoqian He <haoqian.he@smartx.com>
Message-Id: <20250416024729.3289157-4-haoqian.he@smartx.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
This patch captures the error of vhost_virtqueue_stop() in vhost_dev_stop()
and returns the error upward.
Specifically, if QEMU is disconnected from the vhost backend, some actions
in vhost_dev_stop() will fail, such as sending vhost-user messages to the
backend (GET_VRING_BASE, SET_VRING_ENABLE) and vhost_reset_status.
Considering that both set_vring_enable and vhost_reset_status require setting
the specific virtio feature bit, we can capture vhost_virtqueue_stop()'s
error to indicate that QEMU has lost connection with the backend.
This patch is the pre patch for 'vhost-user: return failure if backend crashes
when live migration', which makes the live migration aware of the loss of
connection with the vhost-user backend and aborts the live migration.
Signed-off-by: Haoqian He <haoqian.he@smartx.com>
Message-Id: <20250416024729.3289157-3-haoqian.he@smartx.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
This patch adds the new VM state change cb type `VMChangeStateHandlerWithRet`,
which has return value for `VMChangeStateEntry`.
Thus, we can register a new VM state change cb with return value for device.
Note that `VMChangeStateHandler` and `VMChangeStateHandlerWithRet` are mutually
exclusive and cannot be provided at the same time.
This patch is the pre patch for 'vhost-user: return failure if backend crashes
when live migration', which makes the live migration aware of the loss of
connection with the vhost-user backend and aborts the live migration.
Virtio device will use VMChangeStateHandlerWithRet.
Signed-off-by: Haoqian He <haoqian.he@smartx.com>
Message-Id: <20250416024729.3289157-2-haoqian.he@smartx.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
The pci-testdev device allows for an optional BAR. We have
historically used this without backing to test that systems and OSes
can accomodate large PCI BARs. However to help test p2pdma operations
it is helpful to add an option to back this BAR with host memory.
We add a membar-backed boolean parameter and when set to true or on we
do a host RAM backing. The default is false which ensures backward
compatability.
Signed-off-by: Stephen Bates <sbates@raithlin.com>
Message-Id: <Z_6JhDtn5PlaDgB_@MKMSTEBATES01.amd.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250314-sriov-v9-9-57dae8ae3ab5@daynix.com>
Tested-by: Yui Washizu <yui.washidu@gmail.com>
Tested-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250314-sriov-v9-8-57dae8ae3ab5@daynix.com>
Tested-by: Yui Washizu <yui.washidu@gmail.com>
Tested-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
A virtio-net device can be added as a SR-IOV VF to another virtio-pci
device that will be the PF.
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250314-sriov-v9-7-57dae8ae3ab5@daynix.com>
Tested-by: Yui Washizu <yui.washidu@gmail.com>
Tested-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
Allow user to attach SR-IOV VF to a virtio-pci PF.
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250314-sriov-v9-6-57dae8ae3ab5@daynix.com>
Tested-by: Yui Washizu <yui.washidu@gmail.com>
Tested-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
A user can create a SR-IOV device by specifying the PF with the
sriov-pf property of the VFs. The VFs must be added before the PF.
A user-creatable VF must have PCIDeviceClass::sriov_vf_user_creatable
set. Such a VF cannot refer to the PF because it is created before the
PF.
A PF that user-creatable VFs can be attached calls
pcie_sriov_pf_init_from_user_created_vfs() during realization and
pcie_sriov_pf_exit() when exiting.
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250314-sriov-v9-5-57dae8ae3ab5@daynix.com>
Tested-by: Yui Washizu <yui.washidu@gmail.com>
Tested-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
SR-IOV requires PCI Express.
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250314-sriov-v9-4-57dae8ae3ab5@daynix.com>
Tested-by: Yui Washizu <yui.washidu@gmail.com>
Tested-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
A device cannot be a SR-IOV PF and a VF at the same time.
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250314-sriov-v9-3-57dae8ae3ab5@daynix.com>
Tested-by: Yui Washizu <yui.washidu@gmail.com>
Tested-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
pci_config_get_bar_addr() had a division by vf_stride. vf_stride needs
to be non-zero when there are multiple VFs, but the specification does
not prohibit to make it zero when there is only one VF.
Do not perform the division for the first VF to avoid division by zero.
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250314-sriov-v9-2-57dae8ae3ab5@daynix.com>
Tested-by: Yui Washizu <yui.washidu@gmail.com>
Tested-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
A SR-IOV VF cannot have a ROM BAR.
Co-developed-by: Yui Washizu <yui.washidu@gmail.com>
Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20250314-sriov-v9-1-57dae8ae3ab5@daynix.com>
Tested-by: Yui Washizu <yui.washidu@gmail.com>
Tested-by: Pasha Tatashin <pasha.tatashin@soleen.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
Add serial number parameter in the cxl persistent examples.
Signed-off-by: Yuquan Wang <wangyuquan1236@phytium.com.cn>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-Id: <20250305092501.191929-9-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
1) get alert configuration(Opcode 4201h)
2) set alert configuration(Opcode 4202h)
Signed-off-by: Sweta Kumari <s5.kumari@samsung.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-Id: <20250305092501.191929-7-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
CXL r3.2(8.2.10.9.5.3)
CXL spec 3.2 section 8.2.10.9.5.3 describes media operations commands.
CXL devices supports media operations Sanitize and Write zero command.
Signed-off-by: Vinayak Holikatti <vinayak.kh@samsung.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-Id: <20250305092501.191929-6-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
Move the code of calculation of sanitize duration into function
for usability by other sanitize routines
Estimate times based on:
https://pmem.io/documents/NVDIMM_DSM_Interface-V1.8.pdf
Signed-off-by: Davidlohr Bueso <dave@stgolabs.net>
Signed-off-by: Vinayak Holikatti <vinayak.kh@samsung.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-Id: <20250305092501.191929-5-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
commands cxl r3.2 (8.2.10.9.5.3)
CXL spec 3.2 section 8.2.10.9.5.3 describes media operations commands.
CXL devices supports media operations discovery command.
Signed-off-by: Vinayak Holikatti <vinayak.kh@samsung.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-Id: <20250305092501.191929-4-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
Add Get/Set Response Message Limit commands.
Signed-off-by: Davidlohr Bueso <dave@stgolabs.net>
Reviewed-by: Fan Ni <fan.ni@samsung.com>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-Id: <20250305092501.191929-3-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
As of 3.1 spec, background commands can be canceled with a new
abort command. Implement the support, which is advertised in
the CEL. No ad-hoc context undoing is necessary as all the
command logic of the running bg command is done upon completion.
Arbitrarily, the on-going background cmd will not be aborted if
already at least 85% done;
A mutex is introduced to stabilize mbox request cancel command vs
the timer callback being fired scenarios (as well as reading the
mbox registers). While some operations under critical regions
may be unnecessary (irq notifying, cmd callbacks), this is not
a path where performance is important, so simplicity is preferred.
Tested-by: Ajay Joshi <ajay.opensrc@micron.com>
Reviewed-by: Ajay Joshi <ajay.opensrc@micron.com>
Signed-off-by: Davidlohr Bueso <dave@stgolabs.net>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Message-Id: <20250305092501.191929-2-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
|
|
When only the -kernel parameter is used to load the elf kernel, the initrd
is loaded in the ram. If the initrd size is too large, the loading fails,
resulting in a VM startup failure. This patch first loads initrd near
the kernel.
When the nearby memory space of the kernel is insufficient, it tries to
load it to the starting position of high memory. If there is still not
enough, qemu will report an error and ask the user to increase the memory
space for the virtual machine to boot.
Signed-off-by: Xianglai Li <lixianglai@loongson.cn>
Message-Id: <20250506080946.817092-1-lixianglai@loongson.cn>
Signed-off-by: Song Gao <gaosong@loongson.cn>
|
|
Since memory region iomem supports memory access size with 1/2/4/8,
it can be used for memory region iomem8 and iomem32_high. Now remove
memory region iomem8 and iomem32_high, merge them into iomem together.
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20250507023754.1877445-5-maobibo@loongson.cn>
Signed-off-by: Song Gao <gaosong@loongson.cn>
|
|
The original iomem region only supports 4 bytes access size, set it ok
with 1/2/4/8 bytes. Also unaligned memory access is not supported.
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20250507023754.1877445-4-maobibo@loongson.cn>
Signed-off-by: Song Gao <gaosong@loongson.cn>
|
|
Rename memory region iomem32_low with iomem, also change ops name
as follows:
loongarch_pch_pic_reg32_low_ops --> loongarch_pch_pic_ops
loongarch_pch_pic_low_readw --> loongarch_pch_pic_read
loongarch_pch_pic_low_writew --> loongarch_pch_pic_write
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20250507023754.1877445-3-maobibo@loongson.cn>
Signed-off-by: Song Gao <gaosong@loongson.cn>
|
|
Add trace event trace_loongarch_pch_pic_read(), replaces the following
three events:
trace_loongarch_pch_pic_low_readw()
trace_loongarch_pch_pic_high_readw()
trace_loongarch_pch_pic_readb()
The similiar with write trace event.
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20250507023754.1877445-2-maobibo@loongson.cn>
Signed-off-by: Song Gao <gaosong@loongson.cn>
|
|
Add iomem8 region register write operation emulation in generic write
function loongarch_pch_pic_write(), and use this function for iomem8
region.
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20250507023754.1877445-1-maobibo@loongson.cn>
Signed-off-by: Song Gao <gaosong@loongson.cn>
|
|
Add iomem32_high region register write operation emulation in generic
write function loongarch_pch_pic_write(), and use this function for
iomem32_high region.
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20250507023148.1877287-12-maobibo@loongson.cn>
Signed-off-by: Song Gao <gaosong@loongson.cn>
|