Because floatx80 has an explicit integer bit, this permits some
odd encodings where the integer bit is not set correctly for the
floating point value type. In In Intel terminology the
categories are:
exp == 0, int = 0, mantissa == 0 : zeroes
exp == 0, int = 0, mantissa != 0 : denormals
exp == 0, int = 1 : pseudo-denormals
0 < exp < 0x7fff, int = 0 : unnormals
0 < exp < 0x7fff, int = 1 : normals
exp == 0x7fff, int = 0, mantissa == 0 : pseudo-infinities
exp == 0x7fff, int = 1, mantissa == 0 : infinities
exp == 0x7fff, int = 0, mantissa != 0 : pseudo-NaNs
exp == 0x7fff, int = 1, mantissa == 0 : NaNs
The usual IEEE cases of zero, denormal, normal, inf and NaN are always valid.
x87 permits as input also pseudo-denormals.
m68k permits all those and also pseudo-infinities, pseudo-NaNs and unnormals.
Currently we have an ifdef in floatx80_invalid_encoding() to select
the x86 vs m68k behaviour. Add new floatx80_behaviour flags to
select whether pseudo-NaN and unnormal are valid, and use these
(plus the existing pseudo_inf_valid flag) to decide whether these
encodings are invalid at runtime.
We leave pseudo-denormals as always-valid, since both x86 and m68k
accept them.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250224111524.1101196-8-peter.maydell@linaro.org
Message-id: 20250217125055.160887-6-peter.maydell@linaro.org
The definition of which floatx80 encodings are invalid is
target-specific. Currently we handle this with an ifdef, but we
would like to defer this decision to runtime. In preparation, pass a
float_status argument to floatx80_invalid_encoding().
We will change the implementation from ifdef to looking at
the status argument in the following commit.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250224111524.1101196-7-peter.maydell@linaro.org
In Intel terminology, a floatx80 Infinity with the explicit integer
bit clear is a "pseudo-infinity"; for x86 these are not valid
infinity values. m68k is looser and does not care whether the
Integer bit is set or clear in an infinity.
Move this setting to runtime rather than using an ifdef in
floatx80_is_infinity().
Since this was the last use of the floatx80_infinity global constant,
we remove it and its definition here.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250224111524.1101196-6-peter.maydell@linaro.org
Message-id: 20250217125055.160887-5-peter.maydell@linaro.org
Unlike the other float formats, whether a floatx80 value is
considered to be an Infinity is target-dependent. (On x86 if the
explicit integer bit is clear this is a "pseudo-infinity" and not a
valid infinity; m68k does not care about the value of the integer
bit.)
Currently we select this target-specific logic at compile time with
an ifdef. We're going to want to do this at runtime, so change the
floatx80_is_infinity() function to take a float_status.
This commit doesn't change any logic; we'll do that in the
next commit.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250224111524.1101196-5-peter.maydell@linaro.org
The global const floatx80_infinity is (unlike all the other
float*_infinity values) target-specific, because whether the explicit
Integer bit is set or not varies between m68k and i386. We want to
be able to compile softfloat once for multiple targets, so we can't
continue to use a single global whose value needs to be different
between targets.
Replace the direct uses of floatx80_infinity in target/i386 with
calls to the new floatx80_default_inf() function. Note that because
we can ask the function for either a negative or positive infinity,
we don't need to change the sign of a positive infinity via
floatx80_chs() for the negative-Inf case.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250224111524.1101196-4-peter.maydell@linaro.org
Message-id: 20250217125055.160887-4-peter.maydell@linaro.org
The global const floatx80_infinity is (unlike all the other
float*_infinity values) target-specific, because whether the explicit
Integer bit is set or not varies between m68k and i386. We want to
be able to compile softfloat once for multiple targets, so we can't
continue to use a single global whose value needs to be different
between targets.
Replace the direct uses of floatx80_infinity in target/m68k with
calls to the new floatx80_default_inf() function.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250224111524.1101196-3-peter.maydell@linaro.org
Message-id: 20250217125055.160887-3-peter.maydell@linaro.org
Currently we hardcode at compile time whether the floatx80 default
Infinity value has the explicit integer bit set or not (x86 sets it;
m68k does not). To be able to compile softfloat once for all targets
we'd like to move this setting to runtime.
Define a new FloatX80Behaviour enum which is a set of flags that
define the target's floatx80 handling. Initially we define just one
flag, for whether the default Infinity has the Integer bit set or
not, but we will expand this in future commits to cover the other
floatx80 target specifics that we currently make compile-time
settings.
Define a new function floatx80_default_inf() which returns the
appropriate default Infinity value of the given sign, and use it in
the code that was previously directly using the compile-time constant
floatx80_infinity_{low,high} values when packing an infinity into a
floatx80.
Since floatx80 is highly unlikely to be supported in any new
architecture, and the existing code is generally written as "default
to like x87, with an ifdef for m68k", we make the default value for
the floatx80 behaviour flags be "what x87 does". This means we only
need to change the m68k target to specify the behaviour flags.
(Other users of floatx80 are the Arm NWFPE emulation, which is
obsolete and probably not actually doing the right thing anyway, and
the PPC xsrqpxp insn. Making the default be "like x87" avoids our
needing to review and test for behaviour changes there.)
We will clean up the remaining uses of the floatx80_infinity global
constant in subsequent commits.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250224111524.1101196-2-peter.maydell@linaro.org
Message-id: 20250217125055.160887-2-peter.maydell@linaro.org
Currently if the user requests via -machine dumpdtb=file.dtb that we
dump the DTB, but the machine doesn't have a DTB, we silently ignore
the option. This is confusing to users, and is a legacy of the old
board-specific implementation of the option, where if the execution
codepath didn't go via a call to qemu_fdt_dumpdtb() we would never
handle the option.
Now we handle the option in one place in machine.c, we can provide
the user with a useful message if they asked us to dump a DTB when
none exists. qmp_dumpdtb() already produces this error; remove the
logic in handle_machine_dumpdtb() that was there specifically to
avoid hitting it.
While we're here, beef up the error message a bit with a hint, and
make it consistent about "an FDT" rather than "a FDT". (In the
qmp_dumpdtb() case this needs an ERRP_GUARD to make
error_append_hint() work when the caller passes error_fatal.)
Note that the three places where we might report "doesn't have an
FDT" are hit in different situations:
(1) in handle_machine_dumpdtb(), if CONFIG_FDT is not set: this is
because the QEMU binary was built without libfdt at all. The
build system will not let you build with a machine type that
needs an FDT but no libfdt, so here we know both that the machine
doesn't use FDT and that QEMU doesn't have the support:
(2) in the device_tree-stub.c qmp_dumpdtb(): this is used when
we had libfdt at build time but the target architecture didn't
enable any machines which did "select DEVICE_TREE", so here we
know that the machine doesn't use FDT.
(3) in qmp_dumpdtb(), if current_machine->fdt is NULL all we know
is that this machine never set it. That might be because it doesn't
use FDT, or it might be because the user didn't pass an FDT
on the command line and the machine doesn't autogenerate an FDT.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2733
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20250206151214.2947842-7-peter.maydell@linaro.org
Currently we handle the 'dumpdtb' machine sub-option ad-hoc in every
board model that has an FDT. It's up to the board code to make sure
it calls qemu_fdt_dumpdtb() in the right place.
This means we're inconsistent and often just ignore the user's
command line argument:
* if the board doesn't have an FDT at all
* if the board supports FDT, but there happens not to be one
present (usually because of a missing -fdt option)
This isn't very helpful because it gives the user no clue why their
option was ignored.
However, in order to support the QMP/HMP dumpdtb commands we require
now that every FDT machine stores a pointer to the FDT in
MachineState::fdt. This means we can handle -machine dumpdtb
centrally by calling the qmp_dumpdtb() function, unifying its
handling with the QMP/HMP commands. All the board code calls to
qemu_fdt_dumpdtb() can then be removed.
For this commit we retain the existing behaviour that if there
is no FDT we silently ignore the -machine dumpdtb option.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
The boston machine doesn't set MachineState::fdt to the DTB blob that
it has loaded or created, which means that the QMP/HMP dumpdtb
monitor commands don't work.
Setting MachineState::fdt is easy in the non-FIT codepath: we can
simply do so immediately before loading the DTB into guest memory.
The FIT codepath is a bit more awkward as currently the FIT loader
throws away the memory that the FDT was in after it loads it into
guest memory. So we add a void *pfdt argument to load_fit() for it
to store the FDT pointer into.
There is some readjustment required of the pointer handling in
loader-fit.c, so that it applies 'const' only where it should (e.g.
the data pointer we get back from fdt_getprop() is const, because
it's into the middle of the input FDT data, but the pointer that
fit_load_image_alloc() should not be const, because it's freshly
allocated memory that the caller can change if it likes).
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250206151214.2947842-5-peter.maydell@linaro.org
The function boston_fdt_filter() can return NULL on errors (in which
case it will print an error message). When we call this from the
non-FIT-image codepath, we aren't checking the return value, so we
will plough on with a NULL pointer, and segfault in fdt_totalsize().
Check for errors here.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250206151214.2947842-4-peter.maydell@linaro.org
The openrisc machines don't set MachineState::fdt to point to their
DTB blob. This means that although the command line '-machine
dumpdtb=file.dtb' option works, the equivalent QMP and HMP monitor
commands do not, but instead produce the error "This machine doesn't
have a FDT".
Set MachineState::fdt in openrisc_load_fdt(), when we write it to
guest memory.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250206151214.2947842-3-peter.maydell@linaro.org
In hmp_dumpdtb(), we print a message when the command succeeds. This
message is missing the trailing \n, so the HMP command prompt is
printed immediately after it. We also weren't capitalizing 'DTB', or
quoting the filename in the message. Fix these nits.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20250206151214.2947842-2-peter.maydell@linaro.org
The MMIO region size required to support virtualized environments with
large PCI BAR regions can exceed the hardcoded limit configured in QEMU.
For example, a VM with multiple NVIDIA Grace-Hopper GPUs passed through
requires more MMIO memory than the amount provided by VIRT_HIGH_PCIE_MMIO
(currently 512GB). Instead of updating VIRT_HIGH_PCIE_MMIO, introduce a
new parameter, highmem-mmio-size, that specifies the MMIO size required
to support the VM configuration.
Example usage with 1TB MMIO region size:
-machine virt,gic-version=3,highmem-mmio-size=1T
Signed-off-by: Matthew R. Ochs <mochs@nvidia.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Reviewed-by: Shameer Kolothum <shameerali.kolothum.thodi@huawei.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Nicolin Chen <nicolinc@nvidia.com>
Message-id: 20250221145419.1281890-1-mochs@nvidia.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
When we fill in the SMMUEventInfo for SMMU_EVT_F_CD_FETCH we write
the address into the f_ste_fetch member of the union, but then when
we come to read it back in smmuv3_record_event() we will (correctly)
be using the f_cd_fetch member.
This is more like a cosmetics fix since the f_cd_fetch and f_ste_fetch are
basically the same field since they are in the exact same union with exact
same type, but it's conceptually wrong. Use the correct union member.
Signed-off-by: Nicolin Chen <nicolinc@nvidia.com>
Message-id: 20250220213832.80289-1-nicolinc@nvidia.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
virtio,pc,pci: features, fixes, cleanups
Features:
SR-IOV emulation for pci
virtio-mem-pci support for s390
interleave support for cxl
big endian support for vdpa svq
new QAPI events for vhost-user
Also vIOMMU reset order fixups are in.
Fixes, cleanups all over the place.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# -----BEGIN PGP SIGNATURE-----
#
# iQFDBAABCAAtFiEEXQn9CHHI+FuUyooNKB8NuNKNVGkFAme4b8sPHG1zdEByZWRo
# YXQuY29tAAoJECgfDbjSjVRpHKcIAKPJsVqPdda2dJ7b7FdyRT0Q+uwezXqaGHd4
# 7Lzih1wsxYNkwIAyPtEb76/21qiS7BluqlUCfCB66R9xWjP5/KfvAFj4/r4AEduE
# fxAgYzotNpv55zcRbcflMyvQ42WGiZZHC+o5Lp7vDXUP3pIyHrl0Ydh5WmcD+hwS
# BjXvda58TirQpPJ7rUL+sSfLih17zQkkDcfv5/AgorDy1wK09RBKwMx/gq7wG8yJ
# twy8eBY2CmfmFD7eTM+EKqBD2T0kwLEeLfS/F/tl5Fyg6lAiYgYtCbGLpAmWErsg
# XZvfZmwqL7CNzWexGvPFnnLyqwC33WUP0k0kT88Y5wh3/h98blw=
# =tej8
# -----END PGP SIGNATURE-----
# gpg: Signature made Fri 21 Feb 2025 20:21:31 HKT
# gpg: using RSA key 5D09FD0871C8F85B94CA8A0D281F0DB8D28D5469
# gpg: issuer "mst@redhat.com"
# gpg: Good signature from "Michael S. Tsirkin <mst@kernel.org>" [full]
# gpg: aka "Michael S. Tsirkin <mst@redhat.com>" [full]
# 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
* tag 'for_upstream' of https://git.kernel.org/pub/scm/virt/kvm/mst/qemu: (41 commits)
docs/devel/reset: Document reset expectations for DMA and IOMMU
hw/vfio/common: Add a trace point in vfio_reset_handler
hw/arm/smmuv3: Move reset to exit phase
hw/i386/intel-iommu: Migrate to 3-phase reset
hw/virtio/virtio-iommu: Migrate to 3-phase reset
vhost-user-snd: correct the calculation of config_size
net: vhost-user: add QAPI events to report connection state
hw/virtio/virtio-nsm: Respond with correct length
vdpa: Fix endian bugs in shadow virtqueue
MAINTAINERS: add more files to `vhost`
cryptodev/vhost: allocate CryptoDevBackendVhost using g_mem0()
vhost-iova-tree: Update documentation
vhost-iova-tree, svq: Implement GPA->IOVA & partial IOVA->HVA trees
vhost-iova-tree: Implement an IOVA-only tree
amd_iommu: Use correct bitmask to set capability BAR
amd_iommu: Use correct DTE field for interrupt passthrough
hw/virtio: reset virtio balloon stats on machine reset
mem/cxl_type3: support 3, 6, 12 and 16 interleave ways
hw/mem/cxl_type3: Ensure errp is set on realization failure
hw/mem/cxl_type3: Fix special_ops memory leak on msix_init_exclusive_bar() failure
...
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
To avoid any translation faults, the IOMMUs are expected to be
reset after the devices they protect. Document that we expect
DMA requests to be stopped during the 'enter' or 'hold' phase
while IOMMUs should be reset during the 'exit' phase.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Message-Id: <20250218182737.76722-6-eric.auger@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Currently the iommu may be reset before the devices
it protects. For example this happens with virtio-scsi-pci.
when system_reset is issued from qmp monitor: spurious
"virtio: zero sized buffers are not allowed" warnings can
be observed. This happens because outstanding DMA requests
are still happening while the SMMU gets reset.
This can also happen with VFIO devices. In that case
spurious DMA translation faults can be observed on host.
Make sure the SMMU is reset in the 'exit' phase after
all DMA capable devices have been reset during the 'enter'
or 'hold' phase.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Message-Id: <20250218182737.76722-4-eric.auger@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Currently the IOMMU may be reset before the devices
it protects. For example this happens with virtio devices
but also with VFIO devices. In this latter case this
produces spurious translation faults on host.
Let's use 3-phase reset mechanism and reset the IOMMU on
exit phase after all DMA capable devices have been reset
on 'enter' or 'hold' phase.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Zhenzhong Duan <zhenzhong.duan@intel.com>
Message-Id: <20250218182737.76722-3-eric.auger@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Use virtio_get_config_size() rather than sizeof(struct
virtio_snd_config) for the config_size in the vhost-user-snd frontend.
The frontend shall rely on device features for the size of the device
configuration space. The presence of `controls` in the config space
depends on VIRTIO_SND_F_CTLS according to the specification (v1.3):
`
5.14.4 Device Configuration Layout
...
controls
(driver-read-only) indicates a total number of all available control
elements if VIRTIO_SND_F_CTLS has been negotiated.
`
This fixes an issue introduced by commit ab0c7fb2 ("linux-headers:
update to current kvm/next") in which the optional field `controls` is
added to the virtio_snd_config structure. This breaks vhost-user-device
backends that do not implement the `controls` field.
Fixes: ab0c7fb22b ("linux-headers: update to current kvm/next")
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2805
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Suggested-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Matias Ezequiel Vara Larsen <mvaralar@redhat.com>
Message-Id: <20250217131255.829892-1-mvaralar@redhat.com>
Cc: qemu-stable@nongnu.org
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Dorinda Bassey <dbassey@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The netdev reports NETDEV_VHOST_USER_CONNECTED event when
the chardev is connected, and NETDEV_VHOST_USER_DISCONNECTED
when it is disconnected.
The NETDEV_VHOST_USER_CONNECTED event includes the chardev id.
This allows a system manager like libvirt to detect when the server
fails.
For instance with passt:
{ 'execute': 'qmp_capabilities' }
{ "return": { } }
[killing passt here]
{ "timestamp": { "seconds": 1739538634, "microseconds": 920450 },
"event": "NETDEV_VHOST_USER_DISCONNECTED",
"data": { "netdev-id": "netdev0" } }
[automatic reconnection with reconnect-ms]
{ "timestamp": { "seconds": 1739538638, "microseconds": 354181 },
"event": "NETDEV_VHOST_USER_CONNECTED",
"data": { "netdev-id": "netdev0", "chardev-id": "chr0" } }
Tested-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Message-Id: <20250217092550.1172055-1-lvivier@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
When we return a response packet from NSM, we need to indicate its
length according to the content of the response. Prior to this patch, we
returned the length of the source buffer, which may confuse guest code
that relies on the response size.
Fix it by returning the response payload size instead.
Fixes: bb154e3e0c ("device/virtio-nsm: Support for Nitro Secure Module device")
Reported-by: Vikrant Garg <vikrant1garg@gmail.com>
Signed-off-by: Alexander Graf <graf@amazon.com>
Message-Id: <20250213114541.67515-1-graf@amazon.com>
Reviewed-by: Dorjoy Chowdhury <dorjoychy111@gmail.com>
Fixes: bb154e3e0c ("device/virtio-nsm: Support for Nitro Secure Module device")<br>
Reported-by: Vikrant Garg <vikrant1garg@gmail.com>
Signed-off-by: Alexander Graf <graf@amazon.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Vikrant Garg <vikrant1garg@gmail.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Peter Maydell
Matthew R. OchsandPeter Maydell
Nicolin ChenandPeter Maydell
Stefan Hajnoczi
Eric AugerandMichael S. Tsirkin
Matias Ezequiel Vara LarsenandMichael S. Tsirkin
Laurent VivierandMichael S. Tsirkin
Alexander GrafandMichael S. Tsirkin
Konstantin ShkolnyyandMichael S. Tsirkin