qapi: Back out doc comments added just to please qapi.py
This reverts commit 3313b61's changes to tests/qapi-schema/, except
for tests/qapi-schema/doc-*.
We could keep some of these doc comments to serve as positive test
cases. However, they don't actually add to what we get from doc
comment use in actual schemas, as we we don't test output matches
expectations, and don't systematically cover doc comment features.
Proper positive test coverage would be nice.
qapi: Make doc comments optional where we don't need them
Since we added the documentation generator in commit 3313b61, doc
comments are mandatory. That's a very good idea for a schema that
needs to be documented, but has proven to be annoying for testing.
Make doc comments optional again, but add a new directive
{ 'pragma': { 'doc-required': true } }
to let a QAPI schema require them.
Add test cases for the new pragma directive. While there, plug a
minor hole in includ directive test coverage.
Require documentation in the schemas we actually want documented:
qapi-schema.json and qga/qapi-schema.json.
We could probably make qapi2texi.py cope with incomplete
documentation, but for now, simply make it refuse to run unless the
schema has 'doc-required': true.
qmp: allow setting properties to empty string in qmp-shell
The qmp-shell property parser currently rejects attempts to
set string properties to the empty string eg
(QEMU) migrate-set-parameters tls-hostname=
Error while parsing command line: Expected a key=value pair, got 'tls-hostname='
command format: <command-name> [arg-name1=arg1] ... [arg-nameN=argN]
This is caused by checking the wrong condition after splitting
the parameter on '='. The "partition" method will return "" for
the separator field, if the seperator was not present, so that
is the correct thing to check for malformed syntax.
Peter Maydell [Wed, 15 Mar 2017 18:44:05 +0000 (18:44 +0000)]
Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging
virtio, pc: fixes
Some fixes to fallback from using virtio caching,
pls a minor vm gen id fix.
Signed-off-by: Michael S. Tsirkin <[email protected]>
# gpg: Signature made Wed 15 Mar 2017 17:59:25 GMT
# gpg: using RSA key 0x281F0DB8D28D5469
# gpg: Good signature from "Michael S. Tsirkin <[email protected]>"
# gpg: aka "Michael S. Tsirkin <[email protected]>"
# Primary key fingerprint: 0270 606B 6F3C DF3D 0B17 0970 C350 3912 AFBE 8E67
# Subkey fingerprint: 5D09 FD08 71C8 F85B 94CA 8A0D 281F 0DB8 D28D 5469
* remotes/mst/tags/for_upstream:
virtio-pci: reset modern vq meta data
Revert "virtio: unbreak virtio-pci with IOMMU after caching ring translations"
pci: introduce a bus master container
virtio: validate address space cache during init
virtio: destroy region cache during reset
virtio: guard against NULL pfn
Bugfix: Handle error if VM Generation ID device not present
Jason Wang [Tue, 14 Mar 2017 07:25:05 +0000 (15:25 +0800)]
virtio-pci: reset modern vq meta data
We don't reset proxy->vqs[].{num|desc[]|avail[]|used[]}. This means if
a driver enable the vq without setting vq address after reset. The old
addresses were leaked. Fixing this by resetting modern vq meta data
during device reset.
Jason Wang [Mon, 13 Mar 2017 03:29:57 +0000 (11:29 +0800)]
Revert "virtio: unbreak virtio-pci with IOMMU after caching ring translations"
This reverts commit 96a8821d21411f10d77ea994af369c6e5c35a2cc. Previous patch is a better
solution which does not require a strict order between virtio and IOMMU.
Jason Wang [Mon, 13 Mar 2017 03:29:58 +0000 (11:29 +0800)]
pci: introduce a bus master container
96a8821d2141 ("virtio: unbreak virtio-pci with IOMMU after caching ring
translations") tries to make IOMMU works with virtio memory region
cache, but it requires IOMMU to be created before any virtio
devices. This is sub optimal, fixing this by introduce a bus master
container to make sure address space can be initialized during device
registering, and then we can safely set alias and make
bus_master_enable_region as its subregion during bus master
initialization.
Jason Wang [Wed, 15 Mar 2017 11:48:32 +0000 (19:48 +0800)]
virtio: validate address space cache during init
We don't check the return value of address_space_cache_init(), this
may lead buggy driver use incorrect region caches. Instead of
triggering an assert, catch and warn this early in
virtio_init_region_cache().
Jason Wang [Wed, 15 Mar 2017 11:48:31 +0000 (19:48 +0800)]
virtio: destroy region cache during reset
We don't destroy region cache during reset which can make the maps
of previous driver leaked to a buggy or malicious driver that don't
set vring address before starting to use the device. Fix this by
destroy the region cache during reset and validate it before trying to
see them.
Jason Wang [Wed, 15 Mar 2017 11:48:30 +0000 (19:48 +0800)]
virtio: guard against NULL pfn
To avoid access stale memory region cache after reset, this patch
check the existence of virtqueue pfn for all exported virtqueue access
helpers before trying to use them.
Paolo Bonzini [Wed, 15 Mar 2017 08:16:41 +0000 (09:16 +0100)]
coverity-model: model address_space_read/write
Commit eb7eeb8 ("memory: split address_space_read and
address_space_write", 2015-12-17) made address_space_rw
dispatch to one of address_space_read or address_space_write,
rather than vice versa.
For callers of address_space_read and address_space_write this
causes false positive defects when Coverity sees a length-8 write in
address_space_read and a length-4 (e.g. int*) buffer to read into.
As long as the size of the buffer is okay, this is a false positive.
os: don't corrupt pre-existing memory-backend data with prealloc
When using a memory-backend object with prealloc turned on, QEMU
will memset() the first byte in every memory page to zero. While
this might have been acceptable for memory backends associated
with RAM, this corrupts application data for NVDIMMs.
Instead of setting every page to zero, read the current byte
value and then just write that same value back, so we are not
corrupting the original data. Directly write the value instead
of memset()ing it, since there's no benefit to memset for a
single byte write.
Eduardo Habkost [Mon, 12 Dec 2016 20:49:05 +0000 (18:49 -0200)]
machine: Convert abstract typename on compat_props to subclass names
Original problem description by Greg Kurz:
> Since commit "9a4c0e220d8a hw/virtio-pci: fix virtio
> behaviour", passing -device virtio-blk-pci.disable-modern=off
> has no effect on 2.6 machine types because the internal
> virtio-pci.disable-modern=on compat property always prevail.
The same bug also affects other abstract type names mentioned on
compat_props by machine-types: apic-common, i386-cpu, pci-device,
powerpc64-cpu, s390-skeys, spapr-pci-host-bridge, usb-device,
virtio-pci, x86_64-cpu.
The right fix for this problem is to make sure compat_props and
-global options are always applied in the order they are
registered, instead of reordering them based on the type
hierarchy. But changing the ordering rules of -global is risky
and might break existing configurations, so we shouldn't do that
on a stable branch.
This is a temporary hack that will work around the bug when
registering compat_props properties: if we find an abstract class
on compat_props, register properties for all its non-abstract
subtypes instead. This will make sure -global won't be overridden
by compat_props, while keeping the existing ordering rules on
-global options.
Note that there's one case that won't be fixed by this hack:
"-global spapr-pci-vfio-host-bridge.<option>=<value>" won't be
able to override compat_props, because spapr-pci-host-bridge is
not an abstract class.
* remotes/bonzini/tags/for-upstream:
nbd/client: fix drop_sync [CVE-2017-2630]
memory: info mtree check mr range overflow
icount: process QEMU_CLOCK_VIRTUAL timers in vCPU thread
main-loop: remove now unnecessary optimization
cpus: define QEMUTimerListNotifyCB for QEMU system emulation
qemu-timer: do not include sysemu/cpus.h from util/qemu-timer.h
qemu-timer: fix off-by-one
target/nios2: take BQL around interrupt check
scsi: mptsas: fix the wrong reading size in fetch request
util: Removed unneeded header from path.c
configure: add the missing help output for optional features
scripts/dump-guest-memory.py: fix int128_get64 on recent gcc
kvmclock: Don't crash QEMU if KVM is disabled
kvm: Print MSR information if KVM_{GET,SET}_MSRS failed
exec: add cpu_synchronize_state to cpu_memory_rw_debug
mem-prealloc: reduce large guest start-up and migration time.
docs: Add a note about mixing bootindex with "-boot order"
memory_region: Fix name comments
Peter Maydell [Fri, 3 Mar 2017 11:59:34 +0000 (11:59 +0000)]
hw/misc/imx6_src: Don't crash trying to reset missing CPUs
Commit 4881658a4b introduced a call to arm_get_cpu_by_id(),
and Coverity noticed that we weren't checking that it didn't
return NULL (CID 1371652).
Normally this won't happen (because all 4 CPUs are expected
to exist), but it's possible the user requested fewer CPUs
on the command line. Handle this possibility by silently
doing nothing, which is the same behaviour as before commit 4881658a4b and also how we handle the other CPU operations
(since we ignore the INVALID_PARAM returns from arm_set_cpu_on()
and friends).
There is a slight behavioural difference to the pre-4881658a4b
situation: the "reset this core" bit will remain set rather
than not being permitted to be set. The imx6 datasheet is
unclear about the behaviour in this odd corner case, so we
opt for the simpler code rather than complicated logic to
maintain identical behaviour.
Paolo Bonzini [Thu, 2 Mar 2017 18:56:40 +0000 (19:56 +0100)]
icount: process QEMU_CLOCK_VIRTUAL timers in vCPU thread
icount has become much slower after tcg_cpu_exec has stopped
using the BQL. There is also a latent bug that is masked by
the slowness.
The slowness happens because every occurrence of a QEMU_CLOCK_VIRTUAL
timer now has to wake up the I/O thread and wait for it. The rendez-vous
is mediated by the BQL QemuMutex:
- handle_icount_deadline wakes up the I/O thread with BQL taken
- the I/O thread wakes up and waits on the BQL
- the VCPU thread releases the BQL a little later
- the I/O thread raises an interrupt, which calls qemu_cpu_kick
- the VCPU thread notices the interrupt, takes the BQL to
process it and waits on it
All this back and forth is extremely expensive, causing a 6 to 8-fold
slowdown when icount is turned on.
One may think that the issue is that the VCPU thread is too dependent
on the BQL, but then the latent bug comes in. I first tried removing
the BQL completely from the x86 cpu_exec, only to see everything break.
The only way to fix it (and make everything slow again) was to add a dummy
BQL lock/unlock pair.
This is because in -icount mode you really have to process the events
before the CPU restarts executing the next instruction. Therefore, this
series moves the processing of QEMU_CLOCK_VIRTUAL timers straight in
the vCPU thread when running in icount mode.
The required changes include:
- make the timer notification callback wake up TCG's single vCPU thread
when run from another thread. By using async_run_on_cpu, the callback
can override all_cpu_threads_idle() when the CPU is halted.
- move handle_icount_deadline after qemu_tcg_wait_io_event, so that
the timer notification callback is invoked after the dummy work item
wakes up the vCPU thread
- make handle_icount_deadline run the timers instead of just waking the
I/O thread.
Paolo Bonzini [Fri, 3 Mar 2017 10:51:07 +0000 (11:51 +0100)]
main-loop: remove now unnecessary optimization
This optimization is not necessary anymore, because the vCPU now drops
the I/O thread lock even with TCG. Drop it to simplify the code and
avoid the "I/O thread spun for 1000 iterations" warning.
Paolo Bonzini [Fri, 3 Mar 2017 10:37:57 +0000 (11:37 +0100)]
qemu-timer: fix off-by-one
If the first timer is exactly at the current value of the clock, the
deadline is met and the timer should fire. This fixes itself on the next
iteration of the loop without icount; with icount, however, execution
of instructions will stop exactly at the deadline and won't proceed.
scripts/dump-guest-memory.py: fix int128_get64 on recent gcc
The Int128 is no longer a struct, reaching a python exception:
Python Exception <class 'gdb.error'> Attempt to extract a component of a value that is not a (null).:
Replace struct access with a cast to uint64[] instead.
Eduardo Habkost [Thu, 9 Mar 2017 18:50:46 +0000 (15:50 -0300)]
kvmclock: Don't crash QEMU if KVM is disabled
Most machines don't allow sysbus devices like "kvmclock" to be
created from the command-line, but some of them do (the ones with
has_dynamic_sysbus=true). In those cases, it's possible to
manually create a kvmclock device without KVM being enabled,
making QEMU crash:
Eduardo Habkost [Thu, 9 Mar 2017 19:46:34 +0000 (16:46 -0300)]
kvm: Print MSR information if KVM_{GET,SET}_MSRS failed
When a KVM_{GET,SET}_MSRS ioctl() fails, it is difficult to find
out which MSR caused the problem. Print an error message for
debugging, before we trigger the (ret == cpu->kvm_msr_buf->nmsrs)
assert.
exec: add cpu_synchronize_state to cpu_memory_rw_debug
I sometimes got "Cannot access memory" when using the x command
on the monitor. Turns out that the cpu env did contain stale data
(e.g. wrong control register content for page table origin).
We must synchronize the state of the CPU before walking the page
tables. A similar issues happens for a remote gdb, so lets
do the cpu_synchronize_state in cpu_memory_rw_debug.
Jitendra Kolhe [Fri, 24 Feb 2017 03:31:43 +0000 (09:01 +0530)]
mem-prealloc: reduce large guest start-up and migration time.
Using "-mem-prealloc" option for a large guest leads to higher guest
start-up and migration time. This is because with "-mem-prealloc" option
qemu tries to map every guest page (create address translations), and
make sure the pages are available during runtime. virsh/libvirt by
default, seems to use "-mem-prealloc" option in case the guest is
configured to use huge pages. The patch tries to map all guest pages
simultaneously by spawning multiple threads. Currently limiting the
change to QEMU library functions on POSIX compliant host only, as we are
not sure if the problem exists on win32. Below are some stats with
"-mem-prealloc" option for guest configured to use huge pages.
Changed in v2:
- modify number of memset threads spawned to min(smp_cpus, 16).
- removed 64GB memory restriction for spawning memset threads.
Changed in v3:
- limit number of threads spawned based on
min(sysconf(_SC_NPROCESSORS_ONLN), 16, smp_cpus)
- implement memset thread specific siglongjmp in SIGBUS signal_handler.
Changed in v4
- remove sigsetjmp/siglongjmp and SIGBUS unblock/block for main thread
as main thread no longer touches any pages.
- simplify code my returning memset_thread_failed status from
touch_all_pages.
Thomas Huth [Tue, 28 Feb 2017 17:40:01 +0000 (18:40 +0100)]
docs: Add a note about mixing bootindex with "-boot order"
Occasionally the users try to mix the bootindex properties with the
"-boot order" parameter - and this likely does not give the expected
results. So let's add a proper statement that these two concepts
should not be used together.
The 'name' parameter to memory_region_init_* had been marked as debug
only, however vmstate_region_ram uses it as a parameter to
qemu_ram_set_idstr to set RAMBlock names and these form part of the
migration stream.
In armv8, this register implements more than a single bit, with
fine-grained enables for read access to event counters, cycles
counters, and write access to the software increment. This change
implements those checks using custom access functions for the relevant
registers.
Peter Maydell [Tue, 14 Mar 2017 11:15:00 +0000 (11:15 +0000)]
Merge remote-tracking branch 'remotes/jasowang/tags/net-pull-request' into staging
# gpg: Signature made Tue 14 Mar 2017 07:55:01 GMT
# gpg: using RSA key 0xEF04965B398D6211
# gpg: Good signature from "Jason Wang (Jason Wang on RedHat) <[email protected]>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 215D 46F4 8246 689E C77F 3562 EF04 965B 398D 6211
* remotes/jasowang/tags/net-pull-request:
hw/net: implement MIB counters in mcf_fec driver
COLO-compare: Fix trace_event print bug
e1000e: correctly tear down MSI-X memory regions
Peter Maydell [Tue, 14 Mar 2017 10:13:19 +0000 (10:13 +0000)]
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-2.9-20170314' into staging
ppc patch queue for 2017-03-14
This set has a handful og bugfixes to go into qemu-2.9. This includes
an update to the dtc/libfdt submodule which will fix the build errors
seen on some distributions.
build: include sys/sysmacros.h for major() and minor()
The definition of the major() and minor() macros are moving within glibc to
<sys/sysmacros.h>. Include this header when it is available to avoid the
following sorts of build-stopping messages:
qga/commands-posix.c: In function ‘dev_major_minor’:
qga/commands-posix.c:656:13: error: In the GNU C Library, "major" is defined
by <sys/sysmacros.h>. For historical compatibility, it is
currently defined by <sys/types.h> as well, but we plan to
remove this soon. To use "major", include <sys/sysmacros.h>
directly. If you did not intend to use a system-defined macro
"major", you should undefine it after including <sys/types.h>. [-Werror]
*devmajor = major(st.st_rdev);
^~~~~~~~~~~~~~~~~~~~~~~~~~
qga/commands-posix.c:657:13: error: In the GNU C Library, "minor" is defined
by <sys/sysmacros.h>. For historical compatibility, it is
currently defined by <sys/types.h> as well, but we plan to
remove this soon. To use "minor", include <sys/sysmacros.h>
directly. If you did not intend to use a system-defined macro
"minor", you should undefine it after including <sys/types.h>. [-Werror]
*devminor = minor(st.st_rdev);
^~~~~~~~~~~~~~~~~~~~~~~~~~
The additional include allows the build to complete on Fedora 26 (Rawhide)
with glibc version 2.24.90.
Greg Ungerer [Mon, 13 Mar 2017 04:56:57 +0000 (14:56 +1000)]
hw/net: implement MIB counters in mcf_fec driver
The FEC ethernet hardware module used on ColdFire SoC parts contains a
block of RAM used to maintain hardware counters. This block is accessible
via the usual FEC register address space. There is currently no support
for this in the QEMU mcf_fec driver.
Add support for storing a MIB RAM block, and provide register level
access to it. Also implement a basic set of stats collection functions
to populate MIB data fields.
This support tested running a Linux target and using the net-tools
"ethtool -S" option. As of linux-4.9 the kernels FEC driver makes
accesses to the MIB counters during its initialization (which it never
did before), and so this version of Linux will now fail with the QEMU
error:
qemu: hardware error: mcf_fec_read: Bad address 0x200
Paolo Bonzini [Thu, 9 Mar 2017 11:58:30 +0000 (12:58 +0100)]
e1000e: correctly tear down MSI-X memory regions
MSI-X has been disabled by the time the e1000e device is unrealized, hence
msix_uninit is never called. This causes the object to be leaked, which
shows up as a RAMBlock with empty name when attempting migration.
David Gibson [Tue, 14 Mar 2017 01:24:29 +0000 (12:24 +1100)]
dtc: Update submodule to avoid build errors
The currently included version of the dtc/libfdt submodule has some build
errors on certain distributions (including RHEL7). This is due to some
poorly named macros in libfdt.h; they're designed for use with the sparse
static checker, but use reserved names which conflict with some symbols in
the standard headers.
That's been corrected in upstream dtc, this updates the qemu submodule to
bring the fix to qemu.
David Gibson [Tue, 14 Mar 2017 00:54:17 +0000 (11:54 +1100)]
pseries: Don't expose PCIe extended config space on older machine types
bb9986452 "spapr_pci: Advertise access to PCIe extended config space"
allowed guests to access the extended config space of PCI Express devices
via the PAPR interfaces, even though the paravirtualized bus mostly acts
like plain PCI.
However, that patch enabled access unconditionally, including for existing
machine types, which is an unwise change in behaviour. This patch limits
the change to pseries-2.9 (and later) machine types.
dc0ad84 target/ppc: update overflow flags for add/sub
As for 32-bit ppc target extracting bit 63 for overflow is not correct.
Made it dependent on TARGET_LOG_BITS. This had broken booting MacOS
9.2.1 image
Thomas Huth [Wed, 8 Mar 2017 19:58:43 +0000 (20:58 +0100)]
target/ppc: Fix wrong number of UAMR register
The SPR UAMR has the number 13, and not 12. (Fortunately it seems like
Linux is not using this register yet - only the privileged version with
number 29 ... that's why nobody noticed this problem yet)
Peter Maydell [Mon, 13 Mar 2017 13:16:35 +0000 (13:16 +0000)]
Merge remote-tracking branch 'remotes/ehabkost/tags/x86-pull-request' into staging
x86: Haswell TSX blacklist fix for 2.9
# gpg: Signature made Fri 10 Mar 2017 18:45:08 GMT
# gpg: using RSA key 0x2807936F984DC5A6
# gpg: Good signature from "Eduardo Habkost <[email protected]>"
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF D1AA 2807 936F 984D C5A6
* remotes/ehabkost/tags/x86-pull-request:
i386: Change stepping of Haswell to non-blacklisted value
i386/kvm: Blacklist TSX on known broken hosts
i386: host_vendor_fms() helper function
Kevin Wolf [Thu, 9 Mar 2017 10:49:16 +0000 (11:49 +0100)]
commit: Implement .bdrv_refresh_filename
We want query-block to return the right filename, even if a commit job
put a bdrv_commit_top on top of the actual image format driver. Let
bdrv_commit_top.bdrv_refresh_filename get the filename from its backing
file.
Kevin Wolf [Thu, 9 Mar 2017 10:49:16 +0000 (11:49 +0100)]
mirror: Implement .bdrv_refresh_filename
We want query-block to return the right filename, even if a mirror job
put a bdrv_mirror_top on top of the actual image format driver. Let
bdrv_mirror_top.bdrv_refresh_filename get the filename from its backing
file.
In some cases, bdrv_co_get_block_status() is called recursively for the
whole backing chain. The automatically inserted bdrv_commit_top filter
driver must not stop the recursion, so implement a callback that simply
forwards the request to bs->backing.
Eric Blake [Wed, 8 Mar 2017 21:34:29 +0000 (15:34 -0600)]
migration: Document handling of bdrv_is_allocated() errors
Migration is the only code left in the tree that does not react
to bdrv_is_allocated() failures. But as there is no useful way
to react to the failure, and we are merely skipping unallocated
sectors on success, just document that our choice of handling
is intended.
Eric Blake [Wed, 8 Mar 2017 21:34:28 +0000 (15:34 -0600)]
vvfat: React to bdrv_is_allocated() errors
If bdrv_is_allocated() fails, we should react to that failure.
For 2 of the 3 callers, reporting the error was easy. But in
cluster_was_modified() and its lone caller
get_cluster_count_for_direntry(), it's rather invasive to update
the logic to pass the error back; so there, I went with merely
documenting the issue by changing the return type to bool (in
all likelihood, treating the cluster as modified will then
trigger a read which will also fail, and eventually get to an
error - but given the appalling number of abort() calls in this
code, I'm not making it any worse).
Eric Blake [Wed, 8 Mar 2017 21:34:27 +0000 (15:34 -0600)]
backup: React to bdrv_is_allocated() errors
If bdrv_is_allocated() fails, we should immediately do the backup
error action, rather than attempting backup_do_cow() (although
that will likely fail too).
Eric Blake [Wed, 8 Mar 2017 20:02:16 +0000 (14:02 -0600)]
block: Drop unmaintained 'archipelago' driver
The driver has failed to build since commit da34e65, in qemu 2.6,
due to a missing include of qapi/error.h for error_setg().
Since no one has complained in three releases, it is easier to
remove the dead code than to keep it around, especially since it
is not being built by default and therefore prone to bitrot.
Fam Zheng [Wed, 8 Mar 2017 12:08:14 +0000 (20:08 +0800)]
file-posix: Consider max_segments for BlockLimits.max_transfer
BlockLimits.max_transfer can be too high without this fix, guest will
encounter I/O error or even get paused with werror=stop or rerror=stop. The
cause is explained below.
Linux has a separate limit, /sys/block/.../queue/max_segments, which in
the worst case can be more restrictive than the BLKSECTGET which we
already consider (note that they are two different things). So, the
failure scenario before this patch is:
1) host device has max_sectors_kb = 4096 and max_segments = 64;
2) guest learns max_sectors_kb limit from QEMU, but doesn't know
max_segments;
3) guest issues e.g. a 512KB request thinking it's okay, but actually
it's not, because it will be passed through to host device as an
SG_IO req that has niov > 64;
4) host kernel doesn't like the segmenting of the request, and returns
-EINVAL;
This patch checks the max_segments sysfs entry for the host device and
calculates a "conservative" bytes limit using the page size, which is
then merged into the existing max_transfer limit. Guest will discover
this from the usual virtual block device interfaces. (In the case of
scsi-generic, it will be done in the INQUIRY reply interception in
device model.)
The other possibility is to actually propagate it as a separate limit,
but it's not better. On the one hand, there is a big complication: the
limit is per-LUN in QEMU PoV (because we can attach LUNs from different
host HBAs to the same virtio-scsi bus), but the channel to communicate
it in a per-LUN manner is missing down the stack; on the other hand,
two limits versus one doesn't change much about the valid size of I/O
(because guest has no control over host segmenting).
Also, the idea to fall back to bounce buffering in QEMU, upon -EINVAL,
was explored. Unfortunately there is no neat way to ensure the bounce
buffer is less segmented (in terms of DMA addr) than the guest buffer.
Practically, this bug is not very common. It is only reported on a
Emulex (lpfc), so it's okay to get it fixed in the easier way.
Peter Maydell [Mon, 13 Mar 2017 11:26:36 +0000 (11:26 +0000)]
Merge remote-tracking branch 'remotes/famz/tags/docker-pull-request' into staging
# gpg: Signature made Fri 10 Mar 2017 07:15:38 GMT
# gpg: using RSA key 0xCA35624C6A9171C6
# gpg: Good signature from "Fam Zheng <[email protected]>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 5003 7CB7 9706 0F76 F021 AD56 CA35 624C 6A91 71C6
* remotes/famz/tags/docker-pull-request:
docker/dockerfiles/debian-s390-cross: include clang
tests/docker: support proxy / corporate firewall
Eduardo Habkost [Thu, 9 Mar 2017 18:12:12 +0000 (15:12 -0300)]
i386: Change stepping of Haswell to non-blacklisted value
glibc blacklists TSX on Haswell CPUs with model==60 and
stepping < 4. To make the Haswell CPU model more useful, make
those guests actually use TSX by changing CPU stepping to 4.
Eduardo Habkost [Thu, 9 Mar 2017 18:12:11 +0000 (15:12 -0300)]
i386/kvm: Blacklist TSX on known broken hosts
Some Intel CPUs are known to have a broken TSX implementation. A
microcode update from Intel disabled TSX on those CPUs, but
GET_SUPPORTED_CPUID might be reporting it as supported if the
hosts were not updated yet.
Manually fixup the GET_SUPPORTED_CPUID data to ensure we will
never enable TSX when running on those hosts.
Alex Bennée [Mon, 6 Mar 2017 11:28:48 +0000 (11:28 +0000)]
docker/dockerfiles/debian-s390-cross: include clang
It's a silly little limitation on Shippable that is looks for clang
in the container even though we won't use it. The arm/aarch64 cross
builds inherit this from debian.docker but as we needed to use
debian-testing for this we add it here. We also collapse the update
step into one RUN line to remove and intermediate layer of the docker
build.
Peter Maydell [Thu, 9 Mar 2017 18:53:55 +0000 (18:53 +0000)]
Merge remote-tracking branch 'remotes/stsquad/tags/pull-mttcg-fixups-090317-1' into staging
Fix-ups for MTTCG regressions for 2.9
This is the same as v3 posted a few days ago except with a few extra
Reviewed-by tags added.
# gpg: Signature made Thu 09 Mar 2017 10:45:18 GMT
# gpg: using RSA key 0xFBD0DB095A9E2A44
# gpg: Good signature from "Alex Bennée (Master Work Key) <[email protected]>"
# Primary key fingerprint: 6685 AE99 E751 67BC AFC8 DF35 FBD0 DB09 5A9E 2A44
* remotes/stsquad/tags/pull-mttcg-fixups-090317-1:
hw/intc/arm_gic: modernise the DPRINTF
target/arm/helper: make it clear the EC field is also in hex
target-i386: defer VMEXIT to do_interrupt
target/mips: hold BQL for timer interrupts
translate-all: exit cpu_restore_state early if translating
target/xtensa: hold BQL for interrupt processing
s390x/misc_helper.c: wrap IO instructions in BQL
sparc/sparc64: grab BQL before calling cpu_check_irqs
cpus.c: add additional error_report when !TARGET_SUPPORT_MTTCG
target/i386/cpu.h: declare TCG_GUEST_DEFAULT_MO
vl/cpus: be smarter with icount and MTTCG
Alex Bennée [Thu, 2 Mar 2017 19:08:52 +0000 (19:08 +0000)]
hw/intc/arm_gic: modernise the DPRINTF
While I was debugging the icount issues I realised a bunch of the
messages look quite similar. I've fixed this by including __func__ in
the debug print. At the same time I move the a modern if (GATE) style
printf which ensures the compiler can check for format string errors
even if the code gets optimised away in the non-DEBUG_GIC case.
Alex Bennée [Thu, 2 Mar 2017 14:07:16 +0000 (14:07 +0000)]
target/arm/helper: make it clear the EC field is also in hex
..just like the rest of the displayed ESR register. Otherwise people
might scratch their heads if a not obviously hex number is displayed
for the EC field.
Paolo Bonzini [Tue, 7 Mar 2017 11:37:36 +0000 (12:37 +0100)]
target-i386: defer VMEXIT to do_interrupt
Paths through the softmmu code during code generation now need to be audited
to check for double locking of tb_lock. In particular, VMEXIT can take tb_lock
through cpu_vmexit -> cpu_x86_update_cr4 -> tlb_flush.
To avoid this, split VMEXIT delivery in two parts, similar to what is done with
exceptions. cpu_vmexit only records the VMEXIT exit code and information, and
cc->do_interrupt can then deliver it when it is safe to take the lock.
Alex Bennée [Thu, 2 Mar 2017 10:31:32 +0000 (10:31 +0000)]
translate-all: exit cpu_restore_state early if translating
The translation code uses cpu_ld*_code which can trigger a tlb_fill
which if it fails will erroneously attempts a fault resolution. This
never works during translation as the TB being generated hasn't been
added yet. The target should have checked retaddr before calling
cpu_restore_state but for those that have yet to be fixed we do it
here to avoid a recursive tb_lock() under MTTCG's new locking regime.
Alex Bennée [Wed, 1 Mar 2017 16:22:40 +0000 (16:22 +0000)]
s390x/misc_helper.c: wrap IO instructions in BQL
Helpers that can trigger IO events (including interrupts) need to be
protected by the BQL. I've updated all the helpers that call into an
ioinst_handle_* functions.
Alex Bennée [Wed, 1 Mar 2017 13:20:02 +0000 (13:20 +0000)]
sparc/sparc64: grab BQL before calling cpu_check_irqs
IRQ modification is part of device emulation and should be done while
the BQL is held to prevent races when MTTCG is enabled. This adds
assertions in the hw emulation layer and wraps the calls from helpers
in the BQL.
Alex Bennée [Tue, 28 Feb 2017 14:40:17 +0000 (14:40 +0000)]
cpus.c: add additional error_report when !TARGET_SUPPORT_MTTCG
While we may fail the memory ordering check later that can be
confusing. So in cases where TARGET_SUPPORT_MTTCG has yet to be
defined we should say so specifically.
Alex Bennée [Tue, 28 Feb 2017 14:35:14 +0000 (14:35 +0000)]
target/i386/cpu.h: declare TCG_GUEST_DEFAULT_MO
This suppresses the incorrect warning when forcing MTTCG for x86
guests on x86 hosts. A future patch will still warn when
TARGET_SUPPORT_MTTCG hasn't been defined for the guest (which is still
pending for x86).
Alex Bennée [Mon, 27 Feb 2017 17:09:01 +0000 (17:09 +0000)]
vl/cpus: be smarter with icount and MTTCG
The sense of the test was inverted. Make it simple, if icount is
enabled then we disabled MTTCG by default. If the user tries to force
MTTCG upon us then we tell them "no".
Gerd Hoffmann [Mon, 6 Mar 2017 08:31:51 +0000 (09:31 +0100)]
qxl: clear guest_cursor on QXL_CURSOR_HIDE
Make sure we don't leave guest_cursor pointing into nowhere. This might
lead to (rare) live migration failures, due to target trying to restore
the cursor from the stale pointer.
Peter Maydell [Wed, 8 Mar 2017 09:47:52 +0000 (09:47 +0000)]
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Block layer fixes for 2.9.0-rc0
# gpg: Signature made Tue 07 Mar 2017 14:59:18 GMT
# gpg: using RSA key 0x7F09B272C88F2FD6
# gpg: Good signature from "Kevin Wolf <[email protected]>"
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74 56FE 7F09 B272 C88F 2FD6
* remotes/kevin/tags/for-upstream: (27 commits)
commit: Don't use error_abort in commit_start
block: Don't use error_abort in blk_new_open
sheepdog: Support blockdev-add
qapi-schema: Rename SocketAddressFlat's variant tcp to inet
qapi-schema: Rename GlusterServer to SocketAddressFlat
gluster: Plug memory leaks in qemu_gluster_parse_json()
gluster: Don't duplicate qapi-util.c's qapi_enum_parse()
gluster: Drop assumptions on SocketTransport names
sheepdog: Implement bdrv_parse_filename()
sheepdog: Use SocketAddress and socket_connect()
sheepdog: Report errors in pseudo-filename more usefully
sheepdog: Don't truncate long VDI name in _open(), _create()
sheepdog: Fix snapshot ID parsing in _open(), _create, _goto()
sheepdog: Mark sd_snapshot_delete() lossage FIXME
sheepdog: Fix error handling sd_create()
sheepdog: Fix error handling in sd_snapshot_delete()
sheepdog: Defuse time bomb in sd_open() error handling
block: Fix error handling in bdrv_replace_in_backing_chain()
block: Handle permission errors in change_parent_backing_link()
block: Ignore multiple children in bdrv_check_update_perm()
...
Peter Maydell [Tue, 7 Mar 2017 17:06:48 +0000 (17:06 +0000)]
Merge remote-tracking branch 'remotes/armbru/tags/pull-block-2017-02-28-v4' into staging
block: Command line option -blockdev
# gpg: Signature made Tue 07 Mar 2017 15:07:59 GMT
# gpg: using RSA key 0x3870B400EB918653
# 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-block-2017-02-28-v4: (24 commits)
keyval: Support lists
docs/qapi-code-gen.txt: Clarify naming rules
qapi: Improve how keyval input visitor reports unexpected dicts
block: Initial implementation of -blockdev
qapi: New qobject_input_visitor_new_str() for convenience
keyval: Restrict key components to valid QAPI names
qapi: New parse_qapi_name()
test-qapi-util: New, covering qapi/qapi-util.c
monitor: Assert qmp_schema_json[] is sane
test-visitor-serialization: Pass &error_abort to qobject_from_json()
check-qjson: Test errors from qobject_from_json()
block: More detailed syntax error reporting for JSON filenames
qobject: Propagate parse errors through qobject_from_json()
test-qobject-input-visitor: Abort earlier on bad test input
qjson: Abort earlier on qobject_from_jsonf() misuse
libqtest: Fix qmp() & friends to abort on JSON parse errors
qobject: Propagate parse errors through qobject_from_jsonv()
qapi: Factor out common qobject_input_get_keyval()
qapi: Factor out common part of qobject input visitor creation
test-keyval: Cover use with qobject input visitor
...
Additionally permit non-negative integers as key components. A
dictionary's keys must either be all integers or none. If all keys
are integers, convert the dictionary to a list. The set of keys must
be [0,N].
Examples:
* list.1=goner,list.0=null,list.1=eins,list.2=zwei
is equivalent to JSON [ "null", "eins", "zwei" ]
* a.b.c=1,a.b.0=2
is inconsistent: a.b.c clashes with a.b.0
* list.0=null,list.2=eins,list.2=zwei
has a hole: list.1 is missing
Similar design flaw as for objects: there is no way to denote an empty
list. While interpreting "key absent" as empty list seems natural
(removing a list member from the input string works when there are
multiple ones, so why not when there's just one), it doesn't work:
"key absent" already means "optional list absent", which isn't the
same as "empty list present".
Update the keyval object visitor to use this a.0 syntax in error
messages rather than the usual a[0].
is rejected with "Invalid parameter type for 'aio', expected: string".
To make sense of this, you almost have to translate it into the
equivalent QMP command
Improve the error message to "Parameters 'aio.*' are unexpected".
Take care not to confuse the case "unexpected nested parameters"
(i.e. the object is a QDict or QList) with the case "non-string scalar
parameter". The latter is a misuse of the visitor, and should perhaps
be an assertion. Note that test-qobject-input-visitor exercises this
misuse in test_visitor_in_int_keyval(), test_visitor_in_bool_keyval()
and test_visitor_in_number_keyval().
The JSON argument doesn't exactly blend into the existing option
syntax, so the traditional KEY=VALUE,... syntax is also supported,
using dotted keys to do the nesting:
This does not yet support lists, but that will be addressed shortly.
Note that calling qmp_blockdev_add() (say via qmp_marshal_block_add())
right away would crash. We need to stash the configuration for later
instead. This is crudely done, and bypasses QemuOpts, even though
storing configuration is what QemuOpts is for. Need to revamp option
infrastructure to support QAPI types like BlockdevOptions.
keyval: Restrict key components to valid QAPI names
Until now, key components are separated by '.'. This leaves little
room for evolving the syntax, and is incompatible with the __RFQDN_
prefix convention for downstream extensions.
Since key components will be commonly used as QAPI member names by the
QObject input visitor, we can just as well borrow the QAPI naming
rules here: letters, digits, hyphen and period starting with a letter,
with an optional __RFQDN_ prefix for downstream extensions.
projects / qemu.git / log