Alberto Garcia [Wed, 21 Feb 2018 14:08:49 +0000 (16:08 +0200)]
specs/qcow2: Fix documentation of the compressed cluster descriptor
This patch fixes several mistakes in the documentation of the
compressed cluster descriptor:
1) the documentation claims that the cluster descriptor contains the
number of sectors used to store the compressed data, but what it
actually contains is the number of sectors *minus one* or, in other
words, the number of additional sectors after the first one.
2) the width of the fields is incorrectly specified. The number of bits
used by each field is
x = 62 - (cluster_bits - 8) for the offset field
y = (cluster_bits - 8) for the size field
So the offset field's location is [0, x-1], not [0, x] as stated.
3) the size field does not contain the size of the compressed data,
but rather the number of sectors where that data is stored. The
compressed data starts at the exact point specified in the offset
field and ends when there's enough data to produce a cluster of
decompressed data. Both points can be in the middle of a sector,
allowing several compressed clusters to be stored next to one
another, sharing sectors if necessary.
Anton Nefedov [Wed, 14 Feb 2018 16:09:19 +0000 (19:09 +0300)]
iotest 033: add misaligned write-zeroes test via truncate
This new test case only makes sense for qcow2 while iotest 033 is generic;
however it matches the test purpose perfectly and also 033 contains those
do_test() tricks to pass the alignment, which won't look nice being
duplicated in other tests or moved to the common code.
Anton Nefedov [Wed, 14 Feb 2018 16:09:20 +0000 (19:09 +0300)]
block: fix write with zero flag set and iovector provided
The normal bdrv_co_pwritev() use is either
- BDRV_REQ_ZERO_WRITE clear and iovector provided
- BDRV_REQ_ZERO_WRITE set and iovector == NULL
while
- the flag clear and iovector == NULL is an assertion failure
in bdrv_co_do_zero_pwritev()
- the flag set and iovector provided is in fact allowed
(the flag prevails and zeroes are written)
However the alignment logic does not support the latter case so the padding
areas get overwritten with zeroes.
Currently, general functions like bdrv_rw_co() do provide iovector
regardless of flags. So, keep it supported and use bdrv_co_do_zero_pwritev()
alignment for it which also makes the code a bit more obvious anyway.
Eric Blake [Tue, 13 Feb 2018 20:27:01 +0000 (14:27 -0600)]
block: Drop unused .bdrv_co_get_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Now that all drivers have been updated to provide the
byte-based .bdrv_co_block_status(), we can delete the sector-based
interface.
Eric Blake [Tue, 13 Feb 2018 20:27:00 +0000 (14:27 -0600)]
vvfat: Switch to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the vvfat driver accordingly. Note that we
can rely on the block driver having already clamped limits to our
block size, and simplify accordingly.
Eric Blake [Tue, 13 Feb 2018 20:26:58 +0000 (14:26 -0600)]
vmdk: Switch to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the vmdk driver accordingly. Drop the
now-unused vmdk_find_index_in_cluster().
Also, fix a pre-existing bug: if find_extent() fails (unlikely,
since the block layer did a bounds check), then we must return a
failure, rather than 0.
Eric Blake [Tue, 13 Feb 2018 20:26:57 +0000 (14:26 -0600)]
vdi: Switch to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the vdi driver accordingly. Note that the
TODO is already covered (the block layer guarantees bounds of its
requests), and that we can remove the now-unused s->block_sectors.
Eric Blake [Tue, 13 Feb 2018 20:26:53 +0000 (14:26 -0600)]
qed: Switch to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the qed driver accordingly, taking the opportunity
to inline qed_is_allocated_cb() into its lone caller (the callback
used to be important, until we switched qed to coroutines). There is
no intent to optimize based on the want_zero flag for this format.
Eric Blake [Tue, 13 Feb 2018 20:26:52 +0000 (14:26 -0600)]
qcow2: Switch to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the qcow2 driver accordingly.
For now, we are ignoring the 'want_zero' hint. However, it should
be relatively straightforward to honor the hint as a way to return
larger *pnum values when we have consecutive clusters with the same
data/zero status but which differ only in having non-consecutive
mappings.
Eric Blake [Tue, 13 Feb 2018 20:26:51 +0000 (14:26 -0600)]
qcow: Switch to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the qcow driver accordingly. There is no
intent to optimize based on the want_zero flag for this format.
Eric Blake [Tue, 13 Feb 2018 20:26:50 +0000 (14:26 -0600)]
parallels: Switch to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the parallels driver accordingly. Note that
the internal function block_status() is still sector-based, because
it is still in use by other sector-based functions; but that's okay
because request_alignment is 512 as a result of those functions.
For now, no optimizations are added based on the mapping hint.
Eric Blake [Tue, 13 Feb 2018 20:26:48 +0000 (14:26 -0600)]
iscsi: Switch to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the iscsi driver accordingly. In this case,
it is handy to teach iscsi_co_block_status() to handle a NULL map
and file parameter, even though the block layer passes non-NULL
values, because we also call the function directly. For now, there
are no optimizations done based on the want_zero flag.
We can also make the simplification of asserting that the block
layer passed in aligned values.
Eric Blake [Tue, 13 Feb 2018 20:26:47 +0000 (14:26 -0600)]
iscsi: Switch iscsi_allocmap_update() to byte-based
We are gradually converting to byte-based interfaces, as they are
easier to reason about than sector-based. Convert all uses of
the allocmap (no semantic change). Callers that already had bytes
available are simpler, and callers that now scale to bytes will be
easier to switch to byte-based in the future.
Eric Blake [Tue, 13 Feb 2018 20:26:46 +0000 (14:26 -0600)]
iscsi: Switch cluster_sectors to byte-based
We are gradually converting to byte-based interfaces, as they are
easier to reason about than sector-based. Convert all uses of
the cluster size in sectors, along with adding assertions that we
are not dividing by zero.
Eric Blake [Tue, 13 Feb 2018 20:26:45 +0000 (14:26 -0600)]
gluster: Switch to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the gluster driver accordingly.
In want_zero mode, we continue to report fine-grained hole
information (the caller wants as much mapping detail as possible);
but when not in that mode, the caller prefers larger *pnum and
merely cares about what offsets are allocated at this layer, rather
than where the holes live. Since holes still read as zeroes at
this layer (rather than deferring to a backing layer), we can take
the shortcut of skipping find_allocation(), and merely state that
all bytes are allocated.
We can also drop redundant bounds checks that are already
guaranteed by the block layer.
Eric Blake [Tue, 13 Feb 2018 20:26:44 +0000 (14:26 -0600)]
file-posix: Switch to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the file protocol driver accordingly.
In want_zero mode, we continue to report fine-grained hole
information (the caller wants as much mapping detail as possible);
but when not in that mode, the caller prefers larger *pnum and
merely cares about what offsets are allocated at this layer, rather
than where the holes live. Since holes still read as zeroes at
this layer (rather than deferring to a backing layer), we can take
the shortcut of skipping lseek(), and merely state that all bytes
are allocated.
We can also drop redundant bounds checks that are already
guaranteed by the block layer.
Eric Blake [Tue, 13 Feb 2018 20:26:43 +0000 (14:26 -0600)]
block: Switch passthrough drivers to .bdrv_co_block_status()
We are gradually moving away from sector-based interfaces, towards
byte-based. Update the generic helpers, and all passthrough clients
(blkdebug, commit, mirror, throttle) accordingly.
Eric Blake [Tue, 13 Feb 2018 20:26:42 +0000 (14:26 -0600)]
nvme: Drop pointless .bdrv_co_get_block_status()
Commit bdd6a90 has a bug: drivers should never directly set
BDRV_BLOCK_ALLOCATED, but only io.c should do that (as needed).
Instead, drivers should report BDRV_BLOCK_DATA if it knows that
data comes from this BDS.
But let's look at the bigger picture: semantically, the nvme
driver is similar to the nbd, null, and raw drivers (no backing
file, all data comes from this BDS). But while two of those
other drivers have to supply the callback (null because it can
special-case BDRV_BLOCK_ZERO, raw because it can special-case
a different offset), in this case the block layer defaults are
good enough without the callback at all (similar to nbd).
Eric Blake [Tue, 13 Feb 2018 20:26:41 +0000 (14:26 -0600)]
block: Add .bdrv_co_block_status() callback
We are gradually moving away from sector-based interfaces, towards
byte-based. Now that the block layer exposes byte-based allocation,
it's time to tackle the drivers. Add a new callback that operates
on as small as byte boundaries. Subsequent patches will then update
individual drivers, then finally remove .bdrv_co_get_block_status().
The new code also passes through the 'want_zero' hint, which will
allow subsequent patches to further optimize callers that only care
about how much of the image is allocated (want_zero is false),
rather than full details about runs of zeroes and which offsets the
allocation actually maps to (want_zero is true). As part of this
effort, fix another part of the documentation: the claim in commit 4c41cb4 that BDRV_BLOCK_ALLOCATED is short for 'DATA || ZERO' is a
lie at the block layer (see commit e88ae2264), even though it is
how the bit is computed from the driver layer. After all, there
are intentionally cases where we return ZERO but not ALLOCATED at
the block layer, when we know that a read sees zero because the
backing file is too short. Note that the driver interface is thus
slightly different than the public interface with regards to which
bits will be set, and what guarantees are provided on input.
We also add an assertion that any driver using the new callback will
make progress (the only time pnum will be 0 is if the block layer
already handled an out-of-bounds request, or if there is an error);
the old driver interface did not provide this guarantee, which
could lead to some inf-loops in drastic corner-case failures.
Peter Maydell [Fri, 2 Mar 2018 14:37:10 +0000 (14:37 +0000)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20180302' into staging
target-arm queue:
* implement FCMA and RDM v8.1 and v8.3 instructions
* enable Cortex-M33 v8M core, and provide new mps2-an505 board model
that uses it
* decodetree: Propagate return value from translate subroutines
* xlnx-zynqmp: Implement the RTC device
The integer size check was already outside of the opcode switch;
move the floating-point size check outside as well. Unify the
size vs index adjustment between fp and integer paths.
Peter Maydell [Fri, 2 Mar 2018 10:45:40 +0000 (10:45 +0000)]
mps2-an505: New board model: MPS2 with AN505 Cortex-M33 FPGA image
Define a new board model for the MPS2 with an AN505 FPGA image
containing a Cortex-M33. Since the FPGA images for TrustZone
cores (AN505, and the similar AN519 for Cortex-M23) have a
significantly different layout of devices to the non-TrustZone
images, we use a new source file rather than shoehorning them
into the existing mps2.c.
Peter Maydell [Fri, 2 Mar 2018 10:45:40 +0000 (10:45 +0000)]
hw/arm/iotkit: Model Arm IOT Kit
Model the Arm IoT Kit documented in
http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ecm0601256/index.html
The Arm IoT Kit is a subsystem which includes a CPU and some devices,
and is intended be extended by adding extra devices to form a
complete system. It is used in the MPS2 board's AN505 image for the
Cortex-M33.
Add remaining easy registers to iotkit-secctl:
* NSCCFG just routes its two bits out to external GPIO lines
* BRGINSTAT/BRGINTCLR/BRGINTEN can be dummies, because QEMU's
bus fabric can never report errors
Peter Maydell [Fri, 2 Mar 2018 10:45:39 +0000 (10:45 +0000)]
hw/misc/iotkit-secctl: Add handling for PPCs
The IoTKit Security Controller includes various registers
that expose to software the controls for the Peripheral
Protection Controllers in the system. Implement these.
Peter Maydell [Fri, 2 Mar 2018 10:45:39 +0000 (10:45 +0000)]
hw/misc/iotkit-secctl: Arm IoT Kit security controller initial skeleton
The Arm IoT Kit includes a "security controller" which is largely a
collection of registers for controlling the PPCs and other bits of
glue in the system. This commit provides the initial skeleton of the
device, implementing just the ID registers, and a couple of read-only
read-as-zero registers.
Peter Maydell [Fri, 2 Mar 2018 10:45:39 +0000 (10:45 +0000)]
hw/misc/tz-ppc: Model TrustZone peripheral protection controller
Add a model of the TrustZone peripheral protection controller (PPC),
which is used to gate transactions to non-TZ-aware peripherals so
that secure software can configure them to not be accessible to
non-secure software.
Peter Maydell [Fri, 2 Mar 2018 10:45:38 +0000 (10:45 +0000)]
hw/core/split-irq: Device that splits IRQ lines
In some board or SoC models it is necessary to split a qemu_irq line
so that one input can feed multiple outputs. We currently have
qemu_irq_split() for this, but that has several deficiencies:
* it can only handle splitting a line into two
* it unavoidably leaks memory, so it can't be used
in a device that can be deleted
Implement a qdev device that encapsulates splitting of IRQs, with a
configurable number of outputs. (This is in some ways the inverse of
the TYPE_OR_IRQ device.)
Peter Maydell [Fri, 2 Mar 2018 10:45:38 +0000 (10:45 +0000)]
qdev: Add new qdev_init_gpio_in_named_with_opaque()
The function qdev_init_gpio_in_named() passes the DeviceState pointer
as the opaque data pointor for the irq handler function. Usually
this is what you want, but in some cases it would be helpful to use
some other data pointer.
Add a new function qdev_init_gpio_in_named_with_opaque() which allows
the caller to specify the data pointer they want.
Peter Maydell [Fri, 2 Mar 2018 10:45:38 +0000 (10:45 +0000)]
include/hw/or-irq.h: Add missing include guard
The or-irq.h header file is missing the customary guard against
multiple inclusion, which means compilation fails if it gets
included twice. Fix the omission.
Peter Maydell [Fri, 2 Mar 2018 10:45:38 +0000 (10:45 +0000)]
hw/misc/unimp: Move struct to header file
Move the definition of the struct for the unimplemented-device
from unimp.c to unimp.h, so that users can embed the struct
in their own device structs if they prefer.
Peter Maydell [Fri, 2 Mar 2018 10:45:37 +0000 (10:45 +0000)]
target/arm: Define init-svtor property for the reset secure VTOR value
The Cortex-M33 allows the system to specify the reset value of the
secure Vector Table Offset Register (VTOR) by asserting config
signals. In particular, guest images for the MPS2 AN505 board rely
on the MPS2's initial VTOR being correct for that board.
Implement a QEMU property so board and SoC code can set the reset
value to the correct value.
Peter Maydell [Fri, 2 Mar 2018 10:45:36 +0000 (10:45 +0000)]
armv7m: Forward idau property to CPU object
Create an "idau" property on the armv7m container object which
we can forward to the CPU object. Annoyingly, we can't use
object_property_add_alias() because the CPU object we want to
forward to doesn't exist until the armv7m container is realized.
Peter Maydell [Fri, 2 Mar 2018 10:45:36 +0000 (10:45 +0000)]
target/arm: Define an IDAU interface
In v8M, the Implementation Defined Attribution Unit (IDAU) is
a small piece of hardware typically implemented in the SoC
which provides board or SoC specific security attribution
information for each address that the CPU performs MPU/SAU
checks on. For QEMU, we model this with a QOM interface which
is implemented by the board or SoC object and connected to
the CPU using a link property.
This commit defines the new interface class, adds the link
property to the CPU object, and makes the SAU checking
code call the IDAU interface if one is present.
Peter Maydell [Fri, 2 Mar 2018 10:45:36 +0000 (10:45 +0000)]
hw/arm/armv7m: Honour CPU's address space for image loads
Instead of loading guest images to the system address space, use the
CPU's address space. This is important if we're trying to load the
file to memory or via an alias memory region that is provided by an
SoC object and thus not mapped into the system address space.
Peter Maydell [Fri, 2 Mar 2018 10:45:36 +0000 (10:45 +0000)]
hw/arm/boot: Honour CPU's address space for image loads
Instead of loading kernels, device trees, and the like to
the system address space, use the CPU's address space. This
is important if we're trying to load the file to memory or
via an alias memory region that is provided by an SoC
object and thus not mapped into the system address space.
Peter Maydell [Fri, 2 Mar 2018 10:45:35 +0000 (10:45 +0000)]
loader: Add new load_ramdisk_as()
Add a function load_ramdisk_as() which behaves like the existing
load_ramdisk() but allows the caller to specify the AddressSpace
to use. This matches the pattern we have already for various
other loader functions.
decodetree: Propagate return value from translate subroutines
Allow the translate subroutines to return false for invalid insns.
At present we can of course invoke an invalid insn exception from within
the translate subroutine, but in the short term this consolidates code.
In the long term it would allow the decodetree language to support
overlapping patterns for ISA extensions.
nbd/client: fix error messages in nbd_handle_reply_err
1. NBD_REP_ERR_INVALID is not only about length, so, make message more
general
2. hex format is not very good: it's hard to read something like
"option a (set meta context)", so switch to dec.
Signed-off-by: Vladimir Sementsov-Ogievskiy <[email protected]>
Message-Id: <1518702707[email protected]>
[eblake: expand scope of patch: ALL uses of nbd_opt_lookup and
nbd_rep_lookup are now decimal] Signed-off-by: Eric Blake <[email protected]>
Expose the new constants and structs that will be used by both
server and client implementations of NBD_CMD_BLOCK_STATUS (the
command is currently experimental at
https://github.com/NetworkBlockDevice/nbd/blob/extension-blockstatus/doc/proto.md
but will hopefully be stabilized soon).
Eric Blake [Thu, 15 Feb 2018 03:29:05 +0000 (21:29 -0600)]
nbd: Honor server's advertised minimum block size
Commit 79ba8c98 (v2.7) changed the setting of request_alignment
to occur only during bdrv_refresh_limits(), rather than at at
bdrv_open() time; but at the time, NBD was unaffected, because
it still used sector-based callbacks, so the block layer
defaulted NBD to use 512 request_alignment.
Later, commit 70c4fb26 (also v2.7) changed NBD to use byte-based
callbacks, without setting request_alignment. This resulted in
NBD using request_alignment of 1, which works great when the
server supports it (as is the case for qemu-nbd), but falls apart
miserably if the server requires alignment (but only if qemu
actually sends a sub-sector request; qemu-io can do it, but
most qemu operations still perform on sectors or larger).
Even later, the NBD protocol was updated to document that clients
should learn the server's minimum alignment during NBD_OPT_GO;
and recommended that clients should assume a minimum size of 512
unless the server understands NBD_OPT_GO and replied with a smaller
size. Commit 081dd1fe (v2.10) attempted to do that, by assigning
request_alignment to whatever was learned from the server; but
it has two flaws: the assignment is done during bdrv_open() so
it gets unconditionally wiped out back to 1 during any later
bdrv_refresh_limits(); and the code is not using a default of 512
when the server did not report a minimum size.
Fix these issues by moving the assignment to request_alignment
to the right function, and by using a sane default when the
server does not advertise a minimum size.
Signed-off-by: Michael S. Tsirkin <[email protected]>
# gpg: Signature made Thu 01 Mar 2018 16:27:25 GMT
# gpg: using RSA key 281F0DB8D28D5469
# 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
* remotes/mst/tags/for_upstream:
cryptodev-vhost-user: set the key length
cryptodev-vhost-user: add crypto session handler
cryptodev: add vhost support
cryptodev: add vhost-user as a new cryptodev backend
docs/vmcoreinfo: detail unsupported host format behaviour
vhost: fix incorrect check in vhost_verify_ring_mappings
vhost: avoid to start/stop virtqueue which is not ready
vhost: fix memslot limit check
docs: pcie: Spell out machine type needs for PCIe features
docs: document virtio-balloon stats
intel-iommu: Accept 64-bit writes to FEADDR
virtio-pci: trivial fixes in error message
vhost-user: fix memory leak
Peter Maydell [Thu, 1 Mar 2018 17:08:16 +0000 (17:08 +0000)]
Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20180301-v2' into staging
- add query-cpus-fast and deprecate query-cpus, while adding s390 cpu
information
- remove s390x memory hotplug implementation, which is not useable in
this form
- add boot menu support in the s390-ccw bios
- expose s390x guest crash information
- fixes and cleaups
* remotes/cohuck/tags/s390x-20180301-v2: (27 commits)
s390x/tcg: fix loading 31bit PSWs with the highest bit set
s390x: remove s390_get_memslot_count
s390x/sclp: remove memory hotplug support
s390x/cpumodel: document S390FeatDef.bit not applicable
hmp: change hmp_info_cpus to use query-cpus-fast
qemu-doc: deprecate query-cpus
qmp: add architecture specific cpu data for query-cpus-fast
qmp: add query-cpus-fast
qmp: expose s390-specific CPU info
s390x/tcg: add various alignment checks
s390x/tcg: fix disabling/enabling DAT
s390/stattrib: Make SaveVMHandlers data static
s390x/cpu: expose the guest crash information
pc-bios/s390: Rebuild the s390x firmware images with the boot menu changes
s390-ccw: interactive boot menu for scsi
s390-ccw: use zipl values when no boot menu options are present
s390-ccw: set cp_receive mask only when needed and consume pending service irqs
s390-ccw: read user input for boot index via the SCLP console
s390-ccw: print zipl boot menu
s390-ccw: read stage2 boot loader data to find menu
...
Gonglei [Thu, 1 Mar 2018 13:46:30 +0000 (21:46 +0800)]
cryptodev-vhost-user: add crypto session handler
Introduce two vhost-user meassges: VHOST_USER_CREATE_CRYPTO_SESSION
and VHOST_USER_CLOSE_CRYPTO_SESSION. At this point, the QEMU side
support crypto operation in cryptodev host-user backend.
Gonglei [Thu, 1 Mar 2018 13:46:29 +0000 (21:46 +0800)]
cryptodev: add vhost support
Impliment the vhost-crypto's funtions, such as startup,
stop and notification etc. Introduce an enum
QCryptoCryptoDevBackendOptionsType in order to
identify the cryptodev vhost backend is vhost-user
or vhost-kernel-module (If exist).
Jia He [Wed, 28 Feb 2018 09:35:28 +0000 (01:35 -0800)]
vhost: avoid to start/stop virtqueue which is not ready
In our Armv8a server, we try to configure the vhost scsi but fail
to boot up the guest (-machine virt-2.10). The guest's boot failure
is very early, even earlier than grub.
There are 3 virtqueues (ctrl, event and cmd) for virtio scsi device,
but ovmf and seabios will only set the physical address for the 3rd
one (cmd). Then in vhost_virtqueue_start(), virtio_queue_get_desc_addr
will be 0 for ctrl and event vq when qemu negotiates with ovmf. So
vhost_memory_map fails with ENOMEM.
This patch just fixs it by early quitting the virtqueue start/stop
when virtio_queue_get_desc_addr is 0.
Btw, after guest kernel starts, all the 3 queues will be initialized
and set address correctly.
Peter Maydell [Thu, 1 Mar 2018 15:37:31 +0000 (15:37 +0000)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20180301' into staging
target-arm queue:
* update MAINTAINERS for Alistair's new email address
* add Arm v8.2 FP16 arithmetic extension for linux-user
* implement display connector emulation for vexpress board
* xilinx_spips: Enable only two slaves when reading/writing with stripe
* xilinx_spips: Use 8 dummy cycles with the QIOR/QIOR4 commands
* hw: register: Run post_write hook on reset
* remotes/pmaydell/tags/pull-target-arm-20180301: (42 commits)
MAINTAINERS: Update my email address
linux-user: Report AArch64 FP16 support via hwcap bits
target/arm: Enable ARM_V8_FP16 feature bit for the AArch64 "any" CPU
arm/translate-a64: add all single op FP16 to handle_fp_1src_half
arm/translate-a64: implement simd_scalar_three_reg_same_fp16
arm/translate-a64: add all FP16 ops in simd_scalar_pairwise
arm/translate-a64: add FP16 FMOV to simd_mod_imm
arm/translate-a64: add FP16 FRSQRTE to simd_two_reg_misc_fp16
arm/helper.c: re-factor rsqrte and add rsqrte_f16
arm/translate-a64: add FP16 FSQRT to simd_two_reg_misc_fp16
arm/translate-a64: add FP16 FRCPX to simd_two_reg_misc_fp16
arm/translate-a64: add FP16 FRECPE
arm/helper.c: re-factor recpe and add recepe_f16
arm/translate-a64: add FP16 FNEG/FABS to simd_two_reg_misc_fp16
arm/translate-a64: add FP16 SCVTF/UCVFT to simd_two_reg_misc_fp16
arm/translate-a64: add FP16 FCMxx (zero) to simd_two_reg_misc_fp16
arm/translate-a64: add FCVTxx to simd_two_reg_misc_fp16
arm/translate-a64: add FP16 FPRINTx to simd_two_reg_misc_fp16
arm/translate-a64: initial decode for simd_two_reg_misc_fp16
arm/translate-a64: add FP16 x2 ops for simd_indexed
...
Jay Zhou [Tue, 27 Feb 2018 07:10:04 +0000 (15:10 +0800)]
vhost: fix memslot limit check
Since used_memslots will be updated to the actual value after
registering memory listener for the first time, move the
memslots limit checking to the right place.
s390x/tcg: fix loading 31bit PSWs with the highest bit set
Let's also put the 31-bit hack in front of the REAL MMU, otherwise right
now we get errors when loading a PSW where the highest bit is set (e.g.
via s390-netboot.img). The highest bit is not masked away, therefore we
inject addressing exceptions into the guest.
The proper fix will later be to do all address wrapping before accessing
the MMU - so we won't get any "wrong" entries in there (which makes
flushing also easier). But that will require more work (wrapping in
load_psw, wrapping when incrementing the PC, wrapping every memory
access).
Alex Bennée [Thu, 1 Mar 2018 11:05:55 +0000 (11:05 +0000)]
arm/helper.c: re-factor rsqrte and add rsqrte_f16
Much like recpe the ARM ARM has simplified the pseudo code for the
calculation which is done on a fixed point 9 bit integer maths. So
while adding f16 we can also clean this up to be a little less heavy
on the floating point and just return the fractional part and leave
the calle's to do the final packing of the result.
projects / qemu.git / log