Paolo Bonzini [Wed, 7 Feb 2018 16:29:20 +0000 (17:29 +0100)]
block: early check for blockers on drive-mirror
Even if an op blocker is present for BLOCK_OP_TYPE_MIRROR_SOURCE,
it is checked a bit late and the result is that the target is
created even if drive-mirror subsequently fails. Add an early
check to avoid this.
Alberto Garcia [Fri, 9 Feb 2018 14:42:22 +0000 (16:42 +0200)]
qcow2: Use g_try_realloc() in qcow2_expand_zero_clusters()
g_realloc() aborts the program if it fails to allocate the required
amount of memory. We want to detect that scenario and return an error
instead, so let's use g_try_realloc().
Fam Zheng [Tue, 30 Jan 2018 06:25:03 +0000 (14:25 +0800)]
iotests: Fix CID for VMDK afl image
This reverts commit 76bf133c4 which updated the reference output, and
fixed the reference image, because the code path we want to exercise is
actually the invalid image size.
The descriptor block in the image, which includes the CID to verify, has been
invalid since the reference image was added. Since commit 9877860e7bd we report
this error earlier than the "file too large", so 059.out mismatches.
The binary change is generated along the operations of:
Peter Maydell [Mon, 12 Feb 2018 14:52:48 +0000 (14:52 +0000)]
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-2.12-20180212' into staging
ppc patch queue 2018-02-12
Here's the accumulatead ppc and pseries related patches for the last
while. Highlights are:
* A number of Macintosh / CUDA cleanups from Mark Cave-Ayland
* An important bug fix (missing "break;") for
H_GET_CPU_CHARACTERISTICS
* Yet another fix for SMT mode handling
* Assorted other cleanups and fixes
* remotes/dgibson/tags/ppc-for-2.12-20180212:
misc: introduce new mos6522 VIA device and enable it for ppc builds
cuda: factor out timebase-derived counter value and load time
cuda: set timer 1 frequency property to CUDA_TIMER_FREQ
cuda: don't call cuda_update() when writing to ACR register
cuda: minor cosmetic tidy-ups to get_next_irq_time()
cuda: rename frequency property to tb_frequency
cuda: introduce CUDAState parameter to get_counter()
spapr: set vsmt to MAX(8, smp_threads)
cuda: don't allow writes to port output pins
cuda: do not use old_mmio accesses
hw/ppc: rename functions in comments
spapr: add missing break in h_get_cpu_characteristics()
* remotes/mjt/tags/trivial-patches-fetch:
tests/qapi: use ARRAY_SIZE macro
tests/qapi: use QEMU_IS_ALIGNED macro
tests/hbitmap: use ARRAY_SIZE macro
async: use ARRAY_SIZE macro
qga: use ARRAY_SIZE macro
MAINTAINERS: Add qemu-binfmt-conf.sh script
oslib-posix: check for posix_memalign in configure script
maint: Mention web site maintenance in README
build: fix typo in error message
configure: Allow capstone=git only if git update is not disabled
scripts/make-release: Don't archive .git files
qemu-options.hx: Remove confusing spaces in parameter listings
mailmap: set preferred spelling for Daniel Berrangé
Drop unneeded system header includes
machine: Polish -machine xxx,help
scripts/argparse.py: spelling (independant)
qapi-schema.json: spelling (independant comparation)
misc: introduce new mos6522 VIA device and enable it for ppc builds
The MOS6522 VIA forms the bridge part of several Mac devices, including the
Mac via-cuda and via-pmu devices. Introduce a standard mos6522 device that
can be shared amongst multiple implementations.
This is effectively taking the 6522 parts out of cuda.c and turning them
into a separate device whilst also applying some style tidy-ups and including
a conversion to trace-events.
cuda: factor out timebase-derived counter value and load time
Commit b981289c49 "PPC: Cuda: Use cuda timer to expose tbfreq to guest" altered
the timer calculations from those based upon the hardware CUDA clock frequency
to those based upon the CPU timebase frequency.
In fact we can isolate the differences to 2 simple changes: one to the counter
read value and another to the counter load time. Move these changes into
separate functions so the implementation can be swapped later.
cuda: set timer 1 frequency property to CUDA_TIMER_FREQ
Now that we have successfully decoupled the timebase frequency and the hardware
timer frequency, set the timer 1 frequency property to CUDA_TIMER_FREQ and alter
get_next_irq_time() to use it rather than the hard-coded constant.
In addition to this we must now switch the tb_diff calculation over to use the
timebase frequency now that the hardware clock frequency and the timebase
frequency are different.
Signed-off-by: Mark Cave-Ayland <[email protected]>
[dwg: Correct a conflict due to a bug in an earlier patch] Signed-off-by: David Gibson <[email protected]>
This allows us to more easily differentiate between the timebase frequency used
to calibrate the MacOS timers and the actual frequency of the hardware clock as
indicated by CUDA_TIMER_FREQ.
Signed-off-by: Mark Cave-Ayland <[email protected]> Reviewed-by: Philippe Mathieu-Daudé <[email protected]>
[dwg: Revert some extraneous changes which break compile] Signed-off-by: David Gibson <[email protected]>
Laurent Vivier [Fri, 9 Feb 2018 08:18:58 +0000 (09:18 +0100)]
spapr: set vsmt to MAX(8, smp_threads)
We ignore silently the value of smp_threads when we set
the default VSMT value, and if smp_threads is greater than VSMT
kernel is going into trouble later.
Fixes: 8904e5a750
("spapr: Adjust default VSMT value for better migration compatibility")
oslib-posix: check for posix_memalign in configure script
Check for the presence of posix_memalign() in the configure script,
not using "defined(_POSIX_C_SOURCE) && !defined(__sun__)". This
lets qemu use posix_memalign() on NetBSD versions that have it,
instead of falling back to valloc() which is wasteful when the
required alignment is smaller than a page.
configure: Allow capstone=git only if git update is not disabled
Even with --disable-git-update, ./configure tries updating the capstone
submodule instead of marking it "no"; this disables capstone submodule
if git update is disabled.
Thomas Huth [Tue, 30 Jan 2018 09:36:21 +0000 (10:36 +0100)]
qemu-options.hx: Remove confusing spaces in parameter listings
The spaces between the parameters in the chardev and tpmdev sections
are rather confusing than helpful, and prevent that the lists can be
copy-n-pasted easily for real usage. We also don't use such spaces
in other sections in the documentation, e.g. with the -netdev option,
so let's be consistent and remove the spaces in the chardev and tpmdev
sections, too.
Eric Blake [Wed, 7 Feb 2018 16:03:02 +0000 (10:03 -0600)]
Drop unneeded system header includes
<memory.h> is a non-standard obsolete header that was long ago
replaced by <string.h>.
<malloc.h> is a non-standard header; it is not obsolete (we must
use it for malloc_trim, for example), but generally should not
be used in files that just need malloc() and friends, where
<stdlib.h> is the standard header.
And since osdep.h already guarantees string.h and stdlib.h, we
can drop these unusual system header includes as redundant
rather than replacing them.
Commit bcb5ce08cf ("spapr: Rename machine init functions for clarity")
renamed ppc_spapr_reset to spapr_machine_reset and ppc_spapr_init
to spapr_machine_init. Let's also rename the references in
comments.
We don't need the can_write_zeroes_with_unmap field in
BlockDriverInfo, because it is redundant information with
supported_zero_flags & BDRV_REQ_MAY_UNMAP. Note that
BlockDriverInfo and supported_zero_flags are both per-device
settings, rather than global state about the driver as a
whole, which means one or both of these bits of information
can already be conditional. Let's audit how they were set:
crypto: always setting can_write_ to false is pointless (the
struct starts life zero-initialized), no use of supported_
nbd: just recently fixed to set can_write_ if supported_
includes MAY_UNMAP (thus this commit effectively reverts bca80059e and solves the problem mentioned there in a more
global way)
file-posix, iscsi, qcow2: can_write_ is conditional, while
supported_ was unconditional; but passing MAY_UNMAP would
fail with ENOTSUP if the condition wasn't met
qed: can_write_ is unconditional, but pwrite_zeroes lacks
support for MAY_UNMAP and supported_ is not set. Perhaps
support can be added later (since it would be similar to
qcow2), but for now claiming false is no real loss
all other drivers: can_write_ is not set, and supported_ is
either unset or a passthrough
Simplify the code by moving the conditional into
supported_zero_flags for all drivers, then dropping the
now-unused BDI field. For callers that relied on
bdrv_can_write_zeroes_with_unmap(), we return the same
per-device settings for drivers that had conditions (no
observable change in behavior there); and can now return
true (instead of false) for drivers that support passthrough
(for example, the commit driver) which gives those drivers
the same fix as nbd just got in bca80059e. For callers that
relied on supported_zero_flags, we now have a few more places
that can avoid a wasted call to pwrite_zeroes() that will
just fail with ENOTSUP.
Peter Maydell [Fri, 9 Feb 2018 14:39:09 +0000 (14:39 +0000)]
Merge remote-tracking branch 'remotes/armbru/tags/pull-misc-2018-02-07-v4' into staging
Miscellaneous patches for 2018-02-07
# gpg: Signature made Fri 09 Feb 2018 12:52:51 GMT
# gpg: using RSA key 3870B400EB918653
# gpg: Good signature from "Markus Armbruster <[email protected]>"
# gpg: aka "Markus Armbruster <[email protected]>"
# Primary key fingerprint: 354B C8B3 D7EB 2A6B 6867 4E5F 3870 B400 EB91 8653
* remotes/armbru/tags/pull-misc-2018-02-07-v4:
Move include qemu/option.h from qemu-common.h to actual users
Drop superfluous includes of qapi/qmp/qjson.h
Drop superfluous includes of qapi/qmp/dispatch.h
Include qapi/qmp/qnull.h exactly where needed
Include qapi/qmp/qnum.h exactly where needed
Include qapi/qmp/qbool.h exactly where needed
Include qapi/qmp/qstring.h exactly where needed
Include qapi/qmp/qdict.h exactly where needed
Include qapi/qmp/qlist.h exactly where needed
Include qapi/qmp/qobject.h exactly where needed
qdict qlist: Make most helper macros functions
Eliminate qapi/qmp/types.h
Typedef the subtypes of QObject in qemu/typedefs.h, too
Include qmp-commands.h exactly where needed
Drop superfluous includes of qapi/qmp/qerror.h
Include qapi/error.h exactly where needed
Drop superfluous includes of qapi-types.h and test-qapi-types.h
Clean up includes
Use #include "..." for our own headers, <...> for others
vnc: use stubs for CONFIG_VNC=n dummy functions
Alex Bennée [Wed, 7 Feb 2018 15:19:25 +0000 (15:19 +0000)]
.travis.yml: add --disable-linux-user for some jobs
The modules and co-routine builds are only really relevant to softmmu
builds and regularly timeout on Travis. Let's disable linux-user
builds here for more headroom.
Peter Maydell [Fri, 9 Feb 2018 13:27:40 +0000 (13:27 +0000)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20180209' into staging
target-arm queue:
* Support M profile derived exceptions on exception entry and exit
* Implement AArch64 v8.2 crypto insns (SHA-512, SHA-3, SM3, SM4)
* Implement working i.MX6 SD controller
* Various devices preparatory to i.MX7 support
* Preparatory patches for SVE emulation
* v8M: Fix bug in implementation of 'TT' insn
* Give useful error if user tries to use userspace GICv3 with KVM
* remotes/pmaydell/tags/pull-target-arm-20180209: (30 commits)
hw/core/generic-loader: Allow PC to be set on command line
target/arm/translate.c: Fix missing 'break' for TT insns
target/arm/kvm: gic: Prevent creating userspace GICv3 with KVM
target/arm: Add SVE state to TB->FLAGS
target/arm: Add ZCR_ELx
target/arm: Add SVE to migration state
target/arm: Add predicate registers for SVE
target/arm: Expand vector registers for SVE
hw/arm: Move virt's PSCI DT fixup code to arm/boot.c
usb: Add basic code to emulate Chipidea USB IP
i.MX: Add implementation of i.MX7 GPR IP block
i.MX: Add i.MX7 GPT variant
i.MX: Add code to emulate GPCv2 IP block
i.MX: Add code to emulate i.MX7 SNVS IP-block
i.MX: Add code to emulate i.MX2 watchdog IP block
i.MX: Add code to emulate i.MX7 CCM, PMU and ANALOG IP blocks
hw: i.MX: Convert i.MX6 to use TYPE_IMX_USDHC
sdhci: Add i.MX specific subtype of SDHCI
target/arm: enable user-mode SHA-3, SM3, SM4 and SHA-512 instruction support
target/arm: implement SM4 instructions
...
Move include qemu/option.h from qemu-common.h to actual users
qemu-common.h includes qemu/option.h, but most places that include the
former don't actually need the latter. Drop the include, and add it
to the places that actually need it.
While there, drop superfluous includes of both headers, and
separate #include from file comment with a blank line.
This cleanup makes the number of objects depending on qemu/option.h
drop from 4545 (out of 4743) to 284 in my "build everything" tree.
This cleanup makes the number of objects depending on qapi/qmp/qdict.h
drop from 4550 (out of 4743) to 368 in my "build everything" tree.
For qapi/qmp/qobject.h, the number drops from 4552 to 390.
While there, separate #include from file comment with a blank line.
The macro expansions of qdict_put_TYPE() and qlist_append_TYPE() need
qbool.h, qnull.h, qnum.h and qstring.h to compile. We include qnull.h
and qnum.h in the headers, but not qbool.h and qstring.h. Works,
because we include those wherever the macros get used.
Open-coding these helpers is of dubious value. Turn them into
functions and drop the includes from the headers.
This cleanup makes the number of objects depending on qapi/qmp/qnum.h
from 4551 (out of 4743) to 46 in my "build everything" tree. For
qapi/qmp/qnull.h, the number drops from 4552 to 21.
qapi/qmp/types.h is a convenience header to include a number of
qapi/qmp/ headers. Since we rarely need all of the headers
qapi/qmp/types.h includes, we bypass it most of the time. Most of the
places that use it don't need all the headers, either.
Include the necessary headers directly, and drop qapi/qmp/types.h.
Peter Maydell [Fri, 9 Feb 2018 11:46:32 +0000 (11:46 +0000)]
Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20180209' into staging
s390x updates:
- rework interrupt handling for tcg, smp is now considered non-experimental
- some general improvements in the flic
- improvements in the pci code, and wiring it up in tcg
- add PTFF subfunctions for multiple-epoch to the cpu model
- maintainership updates
- various other fixes and improvements
* remotes/cohuck/tags/s390x-20180209: (29 commits)
MAINTAINERS: add David as additional tcg/s390 maintainer
MAINTAINERS: reorganize s390-ccw bios maintainership
MAINTAINERS: add myself as overall s390x maintainer
s390x/pci: use the right pal and pba in reg_ioat()
s390x/pci: fixup global refresh
s390x/pci: fixup the code walking IOMMU tables
s390x/cpumodel: model PTFF subfunctions for Multiple-epoch facility
s390x/cpumodel: allow zpci features in qemu model
s390x/tcg: wire up pci instructions
s390x/sclp: fix event mask handling
s390x/flic: cache the common flic class in a central function
s390x/kvm: cache the kvm flic in a central function
s390x/tcg: cache the qemu flic in a central function
configure: s390x supports mttcg now
s390x/tcg: remove SMP warning
s390x/tcg: STSI overhaul
s390x: fix size + content of STSI blocks
s390x/flic: optimize CPU wakeup for TCG
s390x/flic: implement qemu_s390_clear_io_flic()
s390x/tcg: implement TEST PENDING INTERRUPTION
...
Peter Maydell [Tue, 6 Feb 2018 10:39:41 +0000 (10:39 +0000)]
target/arm/translate.c: Fix missing 'break' for TT insns
The code where we added the TT instruction was accidentally
missing a 'break', which meant that after generating the code
to execute the TT we would fall through to 'goto illegal_op'
and generate code to take an UNDEF insn.
target/arm/kvm: gic: Prevent creating userspace GICv3 with KVM
KVM doesn't support emulating a GICv3 in userspace, only GICv2. We
currently attempt this anyway, and as a result a KVM guest doesn't
receive interrupts and the user is left wondering why. Report an error
to the user if this particular combination is requested.
Andrey Smirnov [Fri, 9 Feb 2018 10:40:30 +0000 (10:40 +0000)]
hw/arm: Move virt's PSCI DT fixup code to arm/boot.c
Move virt's PSCI DT fixup code to arm/boot.c and set this fixup to
happen automatically for every board that doesn't mark "psci-conduit"
as disabled. This way emulated boards other than "virt" that rely on
PSIC for SMP could benefit from that code.
Peter Maydell [Fri, 9 Feb 2018 10:40:28 +0000 (10:40 +0000)]
target/arm: Make exception vector loads honour the SAU
Make the load of the exception vector from the vector table honour
the SAU and any bus error on the load (possibly provoking a derived
exception), rather than simply aborting if the load fails.
Peter Maydell [Fri, 9 Feb 2018 10:40:27 +0000 (10:40 +0000)]
target/arm: Make v7m_push_callee_stack() honour MPU
Make v7m_push_callee_stack() honour the MPU by using the
new v7m_stack_write() function. We return a flag to indicate
whether the pushes failed, which we can then use in
v7m_exception_taken() to cause us to handle the derived
exception correctly.
Peter Maydell [Fri, 9 Feb 2018 10:40:27 +0000 (10:40 +0000)]
target/arm: Make v7M exception entry stack push check MPU
The memory writes done to push registers on the stack
on exception entry in M profile CPUs are supposed to
go via MPU permissions checks, which may cause us to
take a derived exception instead of the original one of
the MPU lookup fails. We were implementing these as
always-succeeds direct writes to physical memory.
Rewrite v7m_push_stack() to do the necessary checks.
Peter Maydell [Fri, 9 Feb 2018 10:40:27 +0000 (10:40 +0000)]
target/arm: Add ignore_stackfaults argument to v7m_exception_taken()
In the v8M architecture, if the process of taking an exception
results in a further exception this is called a derived exception
(for example, an MPU exception when writing the exception frame to
memory). If the derived exception happens while pushing the initial
stack frame, we must ignore any subsequent possible exception
pushing the callee-saves registers.
In preparation for making the stack writes check for exceptions,
add a return value from v7m_push_stack() and a new parameter to
v7m_exception_taken(), so that the former can tell the latter that
it needs to ignore failures to write to the stack. We also plumb
the argument through to v7m_push_callee_stack(), which is where
the code to ignore the failures will be.
(Note that the v8M ARM pseudocode structures this slightly differently:
derived exceptions cause the attempt to process the original
exception to be abandoned; then at the top level it calls
DerivedLateArrival to prioritize the derived exception and call
TakeException from there. We choose to let the NVIC do the prioritization
and continue forward with a call to TakeException which will then
take either the original or the derived exception. The effect is
the same, but this structure works better for QEMU because we don't
have a convenient top level place to do the abandon-and-retry logic.)
Peter Maydell [Fri, 9 Feb 2018 10:40:27 +0000 (10:40 +0000)]
target/arm: Split "get pending exception info" from "acknowledge it"
Currently armv7m_nvic_acknowledge_irq() does three things:
* make the current highest priority pending interrupt active
* return a bool indicating whether that interrupt is targeting
Secure or NonSecure state
* implicitly tell the caller which is the highest priority
pending interrupt by setting env->v7m.exception
We need to split these jobs, because v7m_exception_taken()
needs to know whether the pending interrupt targets Secure so
it can choose to stack callee-saves registers or not, but it
must not make the interrupt active until after it has done
that stacking, in case the stacking causes a derived exception.
Similarly, it needs to know the number of the pending interrupt
so it can read the correct vector table entry before the
interrupt is made active, because vector table reads might
also cause a derived exception.
Create a new armv7m_nvic_get_pending_irq_info() function which simply
returns information about the highest priority pending interrupt, and
use it to rearrange the v7m_exception_taken() code so we don't
acknowledge the exception until we've done all the things which could
possibly cause a derived exception.
Peter Maydell [Fri, 9 Feb 2018 10:40:27 +0000 (10:40 +0000)]
target/arm: Add armv7m_nvic_set_pending_derived()
In order to support derived exceptions (exceptions generated in
the course of trying to take an exception), we need to be able
to handle prioritizing whether to take the original exception
or the derived exception.
We do this by introducing a new function
armv7m_nvic_set_pending_derived() which the exception-taking code in
helper.c will call when a derived exception occurs. Derived
exceptions are dealt with mostly like normal pending exceptions, so
we share the implementation with the armv7m_nvic_set_pending()
function.
Note that the way we structure this is significantly different
from the v8M Arm ARM pseudocode: that does all the prioritization
logic in the DerivedLateArrival() function, whereas we choose to
let the existing "identify highest priority exception" logic
do the prioritization for us. The effect is the same, though.
Yi Min Zhao [Mon, 5 Feb 2018 07:22:58 +0000 (15:22 +0800)]
s390x/pci: use the right pal and pba in reg_ioat()
When registering ioat, pba should be comprised of leftmost 52 bits and
rightmost 12 binary zeros, and pal should be comprised of leftmost 52
bits and right most 12 binary ones. The lower 12 bits of words 5 and 7
of the FIB are ignored by the facility. Let's fixup this.
Yi Min Zhao [Mon, 5 Feb 2018 07:22:57 +0000 (15:22 +0800)]
s390x/pci: fixup global refresh
The VFIO common code doesn't provide the possibility to modify a
previous mapping entry in another way than unmapping and mapping again
with new properties.
To avoid -EEXIST DMA mapping error, we introduce a GHashTable to store
S390IOTLBEntry instances in order to cache the mapped entries. When
intercepting rpcit instruction, ignore the identical mapped entries to
avoid doing map operations multiple times and do unmap and re-map
operations for the case of updating the valid entries.
Yi Min Zhao [Mon, 5 Feb 2018 07:22:56 +0000 (15:22 +0800)]
s390x/pci: fixup the code walking IOMMU tables
Current s390x PCI IOMMU code is lack of flags' checking, including:
1) protection bit
2) table length
3) table offset
4) intermediate tables' invalid bit
5) format control bit
This patch introduces a new struct named S390IOTLBEntry, and makes up
these missed checkings. At the same time, inform the guest with the
corresponding error number when the check fails. Finally, in order to
get the error number, we export s390_guest_io_table_walk().
s390x/cpumodel: model PTFF subfunctions for Multiple-epoch facility
For now, the kernel does not properly indicate configured CPU subfunctions
to the guest, but simply uses the host values (as support in KVM is still
missing). That's why we missed to model the PTFF subfunctions that come
with Multiple-epoch facility.
Let's properly add these, along with a new feature group.
projects / qemu.git / log