VMSTATE_UNUSED_VARRAY_UINT32 is used to skip a chunk of the stream
that's an n-element array; note the array size and the dynamic value
read never get multiplied so there's no overflow risk.
zhanghailiang [Tue, 17 Jan 2017 12:57:44 +0000 (20:57 +0800)]
COLO: Don't process failover request while loading VM's state
We should not do failover work while the main thread is loading
VM's state. Otherwise the consistent of VM's memory and
device state will be broken.
We will restart the loading process after jump over the stage,
The new failover status 'RELAUNCH' will help to record if we
need to restart the process.
zhanghailiang [Tue, 17 Jan 2017 12:57:43 +0000 (20:57 +0800)]
COLO: Shutdown related socket fd while do failover
If the net connection between primary host and secondary host breaks
while COLO/COLO incoming threads are doing read() or write().
It will block until connection is timeout, and the failover process
will be blocked because of it.
So it is necessary to shutdown all the socket fds used by COLO
to avoid this situation. Besides, we should close the corresponding
file descriptors after failvoer BH shutdown them,
Or there will be an error.
zhanghailiang [Tue, 17 Jan 2017 12:57:42 +0000 (20:57 +0800)]
COLO: fix setting checkpoint-delay not working properly
If we set checkpoint-delay through command 'migrate-set-parameters',
It will not take effect until we finish last sleep chekpoint-delay,
That's will be offensive espeically when we want to change its value
from an extreme big one to a proper value.
Fix it by using timer to realize checkpoint-delay.
Halil Pasic [Fri, 3 Feb 2017 17:52:17 +0000 (18:52 +0100)]
migration: consolidate VMStateField.start
The member VMStateField.start is used for two things, partial data
migration for VBUFFER data (basically provide migration for a
sub-buffer) and for locating next in QTAILQ.
The implementation of the VBUFFER feature is broken when VMSTATE_ALLOC
is used. This however goes unnoticed because actually partial migration
for VBUFFER is not used at all.
Let's consolidate the usage of VMStateField.start by removing support
for partial migration for VBUFFER.
migrate: Introduce zero RAM checks to skip RAM migration
Migration of a "none" machine with no RAM crashes abruptly as
bitmap_new() fails and thus aborts. Instead place zero RAM checks at
appropriate places to skip migration of RAM in this case and complete
migration successfully for devices only.
Pavel Butsykin [Fri, 3 Feb 2017 15:23:21 +0000 (18:23 +0300)]
migration: discard non-dirty ram pages after the start of postcopy
After the start of postcopy migration there are some non-dirty pages which have
already been migrated. These pages are no longer needed on the source vm so that
we can free them and it doen't hurt to complete the migration.
Pavel Butsykin [Fri, 3 Feb 2017 15:23:20 +0000 (18:23 +0300)]
add 'release-ram' migrate capability
This feature frees the migrated memory on the source during postcopy-ram
migration. In the second step of postcopy-ram migration when the source vm
is put on pause we can free unnecessary memory. It will allow, in particular,
to start relaxing the memory stress on the source host in a load-balancing
scenario.
Peter Maydell [Mon, 13 Feb 2017 10:16:23 +0000 (10:16 +0000)]
Merge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2017-02-12' into staging
Block patches
# gpg: Signature made Sun 12 Feb 2017 01:26:20 GMT
# gpg: using RSA key 0xF407DB0061D5CF40
# gpg: Good signature from "Max Reitz <[email protected]>"
# Primary key fingerprint: 91BE B60A 30DB 3E88 57D1 1829 F407 DB00 61D5 CF40
* remotes/maxreitz/tags/pull-block-2017-02-12: (21 commits)
qemu-img: Avoid setting ret to unused value in img_convert()
qemu-img: Use qemu_strtoul() rather than raw strtoul()
qemu-io: don't allow I/O operations larger than BDRV_REQUEST_MAX_BYTES
qcow2: Optimize the refcount-block overlap check
qemu-io: Add failure regression tests
qemu-iotests: Add _unsupported_fmt helper
qemu-io: Return non-zero exit code on failure
block/nfs: fix naming of runtime opts
block/nfs: fix NULL pointer dereference in URI parsing
block: bdrv_invalidate_cache: invalidate children first
block/qapi: reduce the execution time of qmp_query_blockstats
block/qapi: reduce the coupling between the bdrv_query_stats and bdrv_query_bds_stats
qemu-iotest: test to lookup protocol-based image with relative backing
qemu-iotests: Don't create fifos / pidfiles with protocol paths
block: check full backing filename when searching protocol filenames
block/vmdk: Fix the endian problem of buf_len and lba
iotests: record separate timings per format,protocol pair
iotests: Fix reference output for 059
qapi: Tweak error message of bdrv_query_image_info
qemu-img: Improve commit invalid base message
...
Peter Maydell [Mon, 13 Feb 2017 09:30:15 +0000 (09:30 +0000)]
Merge remote-tracking branch 'remotes/awilliam/tags/vfio-updates-20170210.0' into staging
VFIO updates 2017-02-10
- Fix GTT wrap-around for Skylake IGD assignment (Alex Williamson)
- Tag vfio-pci-igd-lpc-bridge as bridge device category (Thomas Huth)
- Don't build calxeda-xgmac or amd-xgbe except on ARM (Thomas Huth)
* remotes/awilliam/tags/vfio-updates-20170210.0:
hw/vfio: Add CONFIG switches for calxeda-xgmac and amd-xgbe
hw/vfio/pci-quirks: Set category of the "vfio-pci-igd-lpc-bridge" device
vfio-pci: Fix GTT wrap-around for Skylake+ IGD
Peter Maydell [Fri, 10 Feb 2017 16:28:24 +0000 (16:28 +0000)]
qemu-img: Avoid setting ret to unused value in img_convert()
Coverity points out that we assign the return value from
bdrv_snapshot_load_tmp() to 'ret' in img_convert(), but then
never use that variable. (We check for failure by looking
at local_err instead.) Drop the unused assignment, bringing
the call into line with the following call to
bdrv_snapshot_laod_tmp_by_id_or_name().
Peter Maydell [Fri, 10 Feb 2017 16:28:23 +0000 (16:28 +0000)]
qemu-img: Use qemu_strtoul() rather than raw strtoul()
Some of the argument parsing in qemu-img uses strtoul() to parse
integer arguments. This is tricky to get correct and in fact the
code does not get it right, because it assigns the result of
strtoul() to an 'int' variable and then tries to check for > INT_MAX.
Coverity correctly complains that the comparison is always false.
Rewrite to use qemu_strtoul(), which has a saner convention for
reporting conversion failures.
Alberto Garcia [Tue, 31 Jan 2017 16:09:54 +0000 (18:09 +0200)]
qemu-io: don't allow I/O operations larger than BDRV_REQUEST_MAX_BYTES
Passing a request size larger than BDRV_REQUEST_MAX_BYTES to any of the
I/O commands results in an error. While 'read' and 'write' handle the
error correctly, 'aio_read' and 'aio_write' hit an assertion:
The reason is that the QEMU I/O code cannot handle request sizes
larger than BDRV_REQUEST_MAX_BYTES, so this patch makes qemu-io check
that all values are within range.
Alberto Garcia [Wed, 1 Feb 2017 12:38:28 +0000 (14:38 +0200)]
qcow2: Optimize the refcount-block overlap check
The metadata overlap checks introduced in a40f1c2add help detect
corruption in the qcow2 image by verifying that data writes don't
overlap with existing metadata sections.
The 'refcount-block' check in particular iterates over the refcount
table in order to get the addresses of all refcount blocks and check
that none of them overlap with the region where we want to write.
The problem with the refcount table is that since it always occupies
complete clusters its size is usually very big. With the default
values of cluster_size=64KB and refcount_bits=16 this table holds 8192
entries, each one of them enough to map 2GB worth of host clusters.
So unless we're using images with several TB of allocated data this
table is going to be mostly empty, and iterating over it is a waste of
CPU. If the storage backend is fast enough this can have an effect on
I/O performance.
This patch keeps the index of the last used (i.e. non-zero) entry in
the refcount table and updates it every time the table changes. The
refcount-block overlap check then uses that index instead of reading
the whole table.
In my tests with a 4GB qcow2 file stored in RAM this doubles the
amount of write IOPS.
Nir Soffer [Wed, 1 Feb 2017 00:31:19 +0000 (02:31 +0200)]
qemu-iotests: Add _unsupported_fmt helper
This helper allows adding tests supporting any format expect the
specified formats. This may be useful to test that many formats behave
in a common way.
Peter Lieven [Wed, 1 Feb 2017 09:53:49 +0000 (10:53 +0100)]
block/nfs: fix naming of runtime opts
commit 94d6a7a accidentally left the naming of runtime opts and QAPI
scheme inconsistent. As one consequence passing of parameters in the
URI is broken. Sync the naming of the runtime opts to the QAPI
scheme.
Please note that this is technically backwards incompatible with the 2.8
release, but the 2.8 release is the only version that had the wrong naming.
Furthermore release 2.8 suffered from a NULL pointer dereference during
URI parsing.
block: bdrv_invalidate_cache: invalidate children first
Current implementation invalidates firstly parent bds and then its
children. This leads to the following bug:
after incoming migration, in bdrv_invalidate_cache_all:
1. invalidate parent bds - reopen it with BDRV_O_INACTIVE cleared
2. child is not yet invalidated
3. parent check that its BDRV_O_INACTIVE is cleared
4. parent writes to child
5. assert in bdrv_co_pwritev, as BDRV_O_INACTIVE is set for child
This patch fixes it by just changing invalidate sequence: invalidate
children first.
Dou Liyang [Sun, 15 Jan 2017 08:01:15 +0000 (16:01 +0800)]
block/qapi: reduce the execution time of qmp_query_blockstats
In order to reduce the execution time, this patch optimize
the qmp_query_blockstats():
Remove the next_query_bds function.
Remove the bdrv_query_stats function.
Remove some judgement sentence.
The original qmp_query_blockstats calls next_query_bds to get
the next objects in each loops. In the next_query_bds, it checks
the query_nodes and blk. It also call bdrv_query_stats to get
the stats, In the bdrv_query_stats, it checks blk and bs each
times. This waste more times, which may stall the main loop a
bit. And if the disk is too many and donot use the dataplane
feature, this may affect the performance in main loop thread.
This patch removes that two functions, and makes the structure
clearly.
Dou Liyang [Sun, 15 Jan 2017 08:01:14 +0000 (16:01 +0800)]
block/qapi: reduce the coupling between the bdrv_query_stats and bdrv_query_bds_stats
The bdrv_query_stats and bdrv_query_bds_stats functions need to call
each other, that increases the coupling. it also makes the program
complicated and makes some unnecessary tests.
Remove the call from bdrv_query_bds_stats to bdrv_query_stats, just
take some recursion to make it clearly.
Avoid testing whether the blk is NULL during querying the bds stats.
It is unnecessary.
Jeff Cody [Thu, 26 Jan 2017 01:08:22 +0000 (20:08 -0500)]
qemu-iotest: test to lookup protocol-based image with relative backing
This test uses NFS and block-stream to force a lookup of a backing
image that has a relative filename, but a full backing image name
with the protocol path intact.
Jeff Cody [Thu, 26 Jan 2017 01:08:20 +0000 (20:08 -0500)]
block: check full backing filename when searching protocol filenames
In bdrv_find_backing_image(), if we are searching an image for a backing
file that contains a protocol, we currently only compare unmodified
paths.
However, some management software will change the backing filename to be
a relative filename in a path. QEMU is able to handle this fine,
because internally it will use path_combine to put together the full
protocol URI.
However, this can lead to an inability to match an image during a QAPI
command that needs to use bdrv_find_backing_image() to find the image,
when it is searched by the full URI.
When searching for a protocol filename, if the straight comparison
fails, this patch will also compare against the full backing filename to
see if that is a match.
QingFeng Hao [Fri, 16 Dec 2016 05:20:40 +0000 (06:20 +0100)]
block/vmdk: Fix the endian problem of buf_len and lba
The problem was triggered by qemu-iotests case 055. It failed when it
was comparing the compressed vmdk image with original test.img.
The cause is that buf_len in vmdk_write_extent wasn't converted to
little-endian before it was stored to disk. But later vmdk_read_extent
read it and converted it from little-endian to cpu endian.
If the cpu is big-endian like s390, the problem will happen and
the data length read by vmdk_read_extent will become invalid!
The fix is to add the conversion in vmdk_write_extent, meanwhile,
repair the endianness problem of lba field which shall also be converted
to little-endian before storing to disk.
iotests: record separate timings per format,protocol pair
The 'check' program records timings for each test that
is run. These timings are only valid, however, for a
particular format/protocol combination. So if frequently
running 'check' with a variety of different formats or
protocols, the times printed can be very misleading.
Instead of having a single 'check.time' file, maintain
multiple 'check.time-$IMGPROTO-$IMGFMT' files.
Max Reitz [Thu, 1 Dec 2016 02:05:08 +0000 (03:05 +0100)]
qemu-img: Improve commit invalid base message
When trying to invoke qemu-img commit with a base image file name that
is not part of the top image's backing chain, the user receives a rather
plain "Base not found" error message. This is not really helpful because
it does not explain what "not found" means, potentially leaving the user
wondering why qemu cannot find a file despite it clearly existing in the
file system.
Improve the error message by clarifying that "not found" means "not
found in the top image's backing chain".
QingFeng Hao [Fri, 16 Dec 2016 05:47:23 +0000 (06:47 +0100)]
iotests: Fix a problem in common.filter
If TEST_DIR is set to /tmp, test case 144 will fail. The reason is that
TEST_DIR resembles 144's test image name tmp.qcow2.
When 144 is testing $TEST_DIR/tmp.qcow2, it wants to replace
$TEST_DIR/tmp.qcow2 to TEST_DIR/tmp.qcow2, but actually it will fail
and get TEST_DIRTEST_DIR.qcow2 in this case.
The fix is just to modify the code to replace $TEST_DIR/ with TEST_DIR/.
Thomas Huth [Fri, 10 Feb 2017 20:12:03 +0000 (13:12 -0700)]
hw/vfio: Add CONFIG switches for calxeda-xgmac and amd-xgbe
Both devices seem to be specific to the ARM platform. It's confusing
for the users if they show up on other target architectures, too
(e.g. when the user runs QEMU with "-device ?" to get a list of
supported devices). Thus let's introduce proper configuration switches
so that the devices are only compiled and included when they are
really required.
Alex Williamson [Fri, 10 Feb 2017 20:12:03 +0000 (13:12 -0700)]
vfio-pci: Fix GTT wrap-around for Skylake+ IGD
Previous IGD, up through Broadwell, only seem to write GTT values into
the first 1MB of space allocated for the BDSM, but clearly the GTT
can be multiple MB in size. Our test in vfio_igd_quirk_data_write()
correctly filters out indexes beyond 1MB, but given the 1MB mask we're
using, we re-apply writes only to the first 1MB of the guest allocated
BDSM.
We can't assume either the host or guest BDSM is naturally aligned, so
we can't simply apply a different mask. Instead, save the host BDSM
and do the arithmetic to subtract the host value to get the BDSM
offset and add it to the guest allocated BDSM.
Peter Maydell [Fri, 10 Feb 2017 18:54:30 +0000 (18:54 +0000)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20170210' into staging
target-arm queue:
* aspeed: minor fixes
* virt: declare fwcfg and virtio-mmio as DMA coherent in DT & ACPI
* arm: enable basic TCG emulation of PMU for AArch64
* remotes/pmaydell/tags/pull-target-arm-20170210:
aspeed/smc: use a modulo to check segment limits
aspeed/smc: handle dummies only in fast read mode
aspeed: remove useless comment on controller segment size
aspeed: check for negative values returned by blk_getlength()
hw/arm/virt: Declare fwcfg as dma cache coherent in dt
hw/arm/virt: Declare fwcfg as dma cache coherent in ACPI
hw/arm/virt: Declare virtio-mmio as dma cache coherent in ACPI
target-arm: Declare virtio-mmio as dma-coherent in dt
target-arm: Enable vPMU support under TCG mode
target-arm: Add support for PMU register PMINTENSET_EL1
target-arm: Add support for AArch64 PMU register PMXEVTYPER_EL0
target-arm: Add support for PMU register PMSELR_EL0
Cédric Le Goater [Fri, 10 Feb 2017 17:40:29 +0000 (17:40 +0000)]
aspeed: remove useless comment on controller segment size
The flash devices used for the FMC controller (BMC firmware) are well
defined for each Aspeed machine and are all smaller than the default
mapping window size, at least for CE0 which is the chip the SoC boots
from.
Cédric Le Goater [Fri, 10 Feb 2017 17:40:29 +0000 (17:40 +0000)]
aspeed: check for negative values returned by blk_getlength()
write_boot_rom() does not check for negative values. This is more a
problem for coverity than the actual code as the size of the flash
device is checked when the m25p80 object is created. If there is
anything wrong with the backing file, we should not even reach that
path.
Alexander Graf [Fri, 10 Feb 2017 17:40:29 +0000 (17:40 +0000)]
hw/arm/virt: Declare fwcfg as dma cache coherent in ACPI
Fw-cfg recently learned how to directly access guest memory and does so in
cache coherent fashion. Tell the guest about that fact when it's using ACPI.
Alexander Graf [Fri, 10 Feb 2017 17:40:29 +0000 (17:40 +0000)]
target-arm: Declare virtio-mmio as dma-coherent in dt
QEMU emulated hardware is always dma coherent with its guest. We do
annotate that correctly on the PCI host controller, but left out
virtio-mmio.
Recent kernels have started to interpret that flag rather than take
dma coherency as granted with virtio-mmio. While that is considered
a kernel bug, as it breaks previously working systems, it showed that
our dt description is incomplete.
This patch adds the respective marker that allows guest OSs to evaluate
that our virtio-mmio devices are indeed cache coherent.
Wei Huang [Fri, 10 Feb 2017 17:40:28 +0000 (17:40 +0000)]
target-arm: Enable vPMU support under TCG mode
This patch contains several fixes to enable vPMU under TCG mode. It
first removes the checking of kvm_enabled() while unsetting
ARM_FEATURE_PMU. With it, the .pmu option can be used to turn on/off vPMU
under TCG mode. Secondly the PMU node of DT table is now created under TCG.
The last fix is to disable the masking of PMUver field of ID_AA64DFR0_EL1.
Wei Huang [Fri, 10 Feb 2017 17:40:28 +0000 (17:40 +0000)]
target-arm: Add support for AArch64 PMU register PMXEVTYPER_EL0
In order to support Linux perf, which uses PMXEVTYPER register,
this patch adds read/write access support for PMXEVTYPER. The access
is CONSTRAINED UNPREDICTABLE when PMSELR is not 0x1f. Additionally
this patch adds support for PMXEVTYPER_EL0.
Ladi Prosek [Fri, 10 Feb 2017 16:47:11 +0000 (11:47 -0500)]
ahci: advertise HOST_CAP_64
The AHCI emulation code supports 64-bit addressing and should advertise this
fact in the Host Capabilities register. Both Linux and Windows drivers test
this bit to decide if the upper 32 bits of various registers may be written
to, and at least some versions of Windows have a bug where DMA is attempted
with an address above 4GB but, in the absence of HOST_CAP_64, the upper 32
bits are left unititialized which leads to a memory corruption.
[Maintainer edit:
This fixes https://bugzilla.redhat.com/show_bug.cgi?id=1411105,
which affects Windows Server 2008 SP2 in some cases.]
Gerd Hoffmann [Thu, 9 Feb 2017 13:02:20 +0000 (14:02 +0100)]
cirrus: fix patterncopy checks
The blit_region_is_unsafe checks don't work correctly for the
patterncopy source. It's a fixed-sized region, which doesn't
depend on cirrus_blt_{width,height}. So go do the check in
cirrus_bitblt_common_patterncopy instead, then tell blit_is_unsafe that
it doesn't need to verify the source. Also handle the case where we
blit from cirrus_bitbuf correctly.
Security impact: I think for the most part error on the safe side this
time, refusing blits which should have been allowed.
Only exception is placing the blit source at the end of the video ram,
so cirrus_blt_srcaddr + 256 goes beyond the end of video memory. But
even in that case I'm not fully sure this actually allows read access to
host memory. To trick the commit 5858dd18 security checks one has to
pick very small cirrus_blt_{width,height} values, which in turn implies
only a fraction of the blit source will actually be used.
Gerd Hoffmann [Mon, 23 Jan 2017 10:26:50 +0000 (11:26 +0100)]
virtio-gpu: fix resource leak in virgl_cmd_resource_unref
When the guest sends VIRTIO_GPU_CMD_RESOURCE_UNREF without detaching the
backing storage beforehand (VIRTIO_GPU_CMD_RESOURCE_DETACH_BACKING)
we'll leak memory.
This patch fixes it for 3d mode, simliar to the 2d mode fix in commit
"b8e2392 virtio-gpu: call cleanup mapping function in resource destroy".
Li Qiang [Sun, 22 Jan 2017 07:42:33 +0000 (23:42 -0800)]
virtio-gpu: fix memory leak in set scanout
In virtio_gpu_set_scanout function, when creating the 'rect'
its refcount is set to 2, by pixman_image_create_bits and
qemu_create_displaysurface_pixman function. This can lead
a memory leak issues. This patch avoid this issue.
Thomas Huth [Fri, 10 Feb 2017 15:14:59 +0000 (16:14 +0100)]
tests/prom-env: Ease time-out problems on slow hosts
Peter Maydell recently ran into time-out problems with the
prom-env test on a rather slow ARM board. To tackle this issue,
we can speed up the test by running QEMU with "-nodefaults" for
the pseries machine, so that SLOF has less devices to scan during
boot, and by using the "nvramrc" environment variable instead of
"boot-command", since this variable is evaluated earlier in the
boot process.
And to be really sure that we do not face such time out problems
again, let's also increase the time out value from 100s to 120s
instead.
Alex Bennée [Fri, 20 Jan 2017 17:11:49 +0000 (17:11 +0000)]
.travis.yml: split VM based builds
The Trusty based builds run a little slower than the main container
based ones. This is also true for the latest version of Clang. The
builds are getting very close (and occasionally run over) the 50 minute
timeout. Rather than partitioning by target I just split them into
linux-user and system builds.
* remotes/kraxel/tags/pull-ui-20170209-2:
ui: add ability to specify multiple VNC listen addresses
util: add iterators for QemuOpts values
ui: let VNC server listen on all resolved IP addresses
ui: extract code to connect/listen from vnc_display_open
ui: refactor code for populating SocketAddress from vnc_display_open
ui: refactor VncDisplay to allow multiple listening sockets
ui: fix reporting of VNC auth in query-vnc-servers
ui: fix regression handling bare 'websocket' option to -vnc
vnc: do not disconnect on EAGAIN
ui/vnc: Drop unused vnc_has_job() and vnc_jobs_clear()
ui: add ability to specify multiple VNC listen addresses
This change allows the listen address and websocket address
options for -vnc to be repeated. This causes the VNC server
to listen on multiple addresses. e.g.
To iterate over all QemuOpts currently requires using a callback
function which is inconvenient for control flow. Add support for
using iterator functions more directly
ui: let VNC server listen on all resolved IP addresses
Remove the limitation that the VNC server can only listen on
a single resolved IP address. This uses the new DNS resolver
API to resolve a SocketAddress struct into an array of
SocketAddress structs containing raw IP addresses. The VNC
server will then attempt to listen on all resolved IP addresses.
The server must successfully listen on at least one of the
resolved IP addresses, otherwise an error will be reported.
ui: extract code to connect/listen from vnc_display_open
The code which takes a SocketAddress and connects/listens on the
network is going to get more complicated to deal with multiple
listeners. Pull it out into a separate method to avoid making the
vnc_display_open method even more complex.
ui: refactor code for populating SocketAddress from vnc_display_open
The code which interprets the CLI args to populate the SocketAddress
objects for plain & websockets VNC is quite complex already and will
need further enhancements shortly. Refactor it into separate methods
to avoid vnc_display_open getting even larger. As a side effect of
the refactoring, it is now possible to specify a listen address for
the websocket server explicitly. e.g,
-vnc localhost:5900,websockets=0.0.0.0:8080
will listen on localhost for the plain VNC server, but expose the
websockets VNC server on the public interface. This refactoring
also removes the restriction that prevents enabling websockets
when the plain VNC server is listening on a UNIX socket.
ui: refactor VncDisplay to allow multiple listening sockets
Currently there is only a single listener for plain VNC and
a single listener for websockets VNC. This means that if
getaddrinfo() returns multiple IP addresses, for a hostname,
the VNC server can only listen on one of them. This is
just bearable if listening on wildcard interface, or if
the host only has a single network interface to listen on,
but if there are multiple NICs and the VNC server needs
to listen on 2 or more specific IP addresses, it can't be
done.
This refactors the VncDisplay state so that it holds an
array of listening sockets, but still only listens on
one socket.
ui: fix reporting of VNC auth in query-vnc-servers
Currently the VNC authentication info is emitted at the
top level of the query-vnc-servers data. This is wrong
because the authentication scheme differs between plain
and websockets when TLS is enabled. We should instead
report auth against the individual servers. e.g.
This also future proofs the QMP schema so that we can
cope with multiple VNC server instances, listening on
different interfaces or ports, with different auth
setup.
ui: fix regression handling bare 'websocket' option to -vnc
The -vnc argument is documented as accepting two syntaxes for
the 'websocket' option, either a bare option name, or a port
number. If using the bare option name, it is supposed to apply
the display number as an offset to base port 5700. e.g.
-vnc localhost:3,websocket
should listen on port 5703, however, this was broken in 2.3.0 since
Michael Tokarev [Fri, 3 Feb 2017 09:52:29 +0000 (12:52 +0300)]
vnc: do not disconnect on EAGAIN
When qemu vnc server is trying to send large update to clients,
there might be a situation when system responds with something
like EAGAIN, indicating that there's no system memory to send
that much data (depending on the network speed, client and server
and what is happening). In this case, something like this happens
on qemu side (from strace):
qemu closes the socket before the retry, and obviously it gets EBADF
when trying to send to -1.
This is because there WAS a special handling for EAGAIN, but now it doesn't
work anymore, after commit 04d2529da27db512dcbd5e99d0e26d333f16efcc, because
now in all error-like cases we initiate vnc disconnect.
This change were introduced in qemu 2.6, and caused numerous grief for many
people, resulting in their vnc clients reporting sporadic random disconnects
from vnc server.
Fix that by doing the disconnect only when necessary, i.e. omitting this
very case of EAGAIN.
Hopefully the existing condition (comparing with QIO_CHANNEL_ERR_BLOCK)
is sufficient, as the original code (before the above commit) were
checking for other errno values too.
Apparently there's another (semi?)bug exist somewhere here, since the
code tries to write to fd# -1, it probably should check if the connection
is open before. But this isn't important.
Peter Maydell [Tue, 7 Feb 2017 19:21:30 +0000 (19:21 +0000)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20170207-1' into staging
target-arm:
* new "unimplemented" device for stubbing out devices in a
system model so accesses can be logged
* stellaris: document the SoC memory map
* arm: create instruction syndromes for AArch32 data aborts
* arm: Correctly handle watchpoints for BE32 CPUs
* Fix Thumb-1 BE32 execution and disassembly
* arm: Add cfgend parameter for ARM CPU selection
* sd: sdhci: check data length during dma_memory_read
* aspeed: add a watchdog controller
* integratorcp: adding vmstate for save/restore
* remotes/pmaydell/tags/pull-target-arm-20170207-1:
stellaris: Use the 'unimplemented' device for parts we don't implement
hw/misc: New "unimplemented" sysbus device
stellaris: Document memory map and which SoC devices are unimplemented
target/arm: A32, T32: Create Instruction Syndromes for Data Aborts
target/arm: Abstract out pbit/wbit tests in ARM ldr/str decode
arm: Correctly handle watchpoints for BE32 CPUs
Fix Thumb-1 BE32 execution and disassembly.
target/arm: Add cfgend parameter for ARM CPU selection.
hw/arm/integratorcp: Support specifying features via -cpu
sd: sdhci: check data length during dma_memory_read
aspeed: add a watchdog controller
wdt: Add Aspeed watchdog device model
integratorcp: adding vmstate for save/restore
Peter Maydell [Tue, 7 Feb 2017 18:30:00 +0000 (18:30 +0000)]
hw/misc: New "unimplemented" sysbus device
Create a new "unimplemented" sysbus device, which simply accepts
all read and write accesses, and implements them as read-as-zero,
write-ignored, with logging of the access as LOG_UNIMP.
This is useful for stubbing out bits of an SoC or board model
which haven't been written yet.
Peter Maydell [Tue, 7 Feb 2017 18:30:00 +0000 (18:30 +0000)]
target/arm: A32, T32: Create Instruction Syndromes for Data Aborts
Add support for generating the ISS (Instruction Specific Syndrome)
for Data Abort exceptions taken from AArch32. These syndromes are
used by hypervisors for example to trap and emulate memory accesses.
This is the equivalent for AArch32 guests of the work done for AArch64
guests in commit aaa1f954d4cab243.
Peter Maydell [Tue, 7 Feb 2017 18:29:59 +0000 (18:29 +0000)]
target/arm: Abstract out pbit/wbit tests in ARM ldr/str decode
In the ARM ldr/str decode path, rather than directly testing
"insn & (1 << 21)" and "insn & (1 << 24)", abstract these
bits out into wbit and pbit local flags. (We will want to
do more tests against them to determine whether we need to
provide syndrome information.)
Julian Brown [Tue, 7 Feb 2017 18:29:59 +0000 (18:29 +0000)]
arm: Correctly handle watchpoints for BE32 CPUs
In BE32 mode, sub-word size watchpoints can fail to trigger because the
address of the access is adjusted in the opcode helpers before being
compared with the watchpoint registers. This patch reverses the address
adjustment before performing the comparison with the help of a new CPUClass
hook.
This version of the patch augments and tidies up comments a little.
Julian Brown [Tue, 7 Feb 2017 18:29:59 +0000 (18:29 +0000)]
Fix Thumb-1 BE32 execution and disassembly.
Thumb-1 code has some issues in BE32 mode (as currently implemented). In
short, since bytes are swapped within words at load time for BE32
executables, this also swaps pairs of adjacent Thumb-1 instructions.
This patch un-swaps those pairs of instructions again, both for execution,
and for disassembly. (The previous version of the patch always read four
bytes in arm_read_memory_func and then extracted the proper two bytes,
in a probably misguided attempt to match the behaviour of actual hardware
as described by e.g. the ARM9TDMI TRM, section 3.3 "Endian effects for
instruction fetches". It's less complicated to just read the correct
two bytes though.)
Julian Brown [Tue, 7 Feb 2017 18:29:59 +0000 (18:29 +0000)]
target/arm: Add cfgend parameter for ARM CPU selection.
Add a new "cfgend" property which selects whether the CPU resets into
big-endian mode or not. This setting affects whether we reset with
SCTLR_B (ARMv6 and earlier) or SCTLR_EE (ARMv7 and later) set.
Signed-off-by: Julian Brown <[email protected]>
Message-id: 11420d1c49636c1790e60578ee996e51f0f0b835.1484929304[email protected]
[PMM: use error_report_err() rather than error_report();
move the integratorcp changes to their own patch;
drop an unnecessary extra #include;
rephrase commit message accordingly;
move setting of reset_sctlr above registration of cpregs
so it actually has an effect] Reviewed-by: Peter Maydell <[email protected]> Signed-off-by: Peter Maydell <[email protected]>
Julian Brown [Tue, 7 Feb 2017 18:29:59 +0000 (18:29 +0000)]
hw/arm/integratorcp: Support specifying features via -cpu
Since the integratorcp board creates the CPU object directly
rather than via cpu_arm_init(), we have to call the CPU
class parse_features() method ourselves if we want to
support the user passing features via the -cpu command
line argument as well as just the cpu name. Do so.
sd: sdhci: check data length during dma_memory_read
While doing multi block SDMA transfer in routine
'sdhci_sdma_transfer_multi_blocks', the 's->fifo_buffer' starting
index 'begin' and data length 's->data_count' could end up to be same.
This could lead to an OOB access issue. Correct transfer data length
to avoid it.
The Aspeed SoC includes a set of watchdog timers using 32-bit
decrement counters, which can be based either on the APB clock or
a 1 MHz clock.
The watchdog timer is designed to prevent system deadlock and, in
general, it should be restarted before timeout. When a timeout occurs,
different types of signals can be generated, ARM reset, SOC reset,
System reset, CPU Interrupt, external signal or boot from alternate
block. The current model only performs the system reset function as
this is used by U-Boot and Linux.
Signed-off-by: Joel Stanley <[email protected]>
Message-id: 1485452251[email protected]
[clg: - fixed compile breakage
- fixed io region size
- added watchdog_perform_action() on timer expiry
- wrote a commit log
- merged fixes from Andrew Jeffery to scale the reload value ] Signed-off-by: Cédric Le Goater <[email protected]> Reviewed-by: Peter Maydell <[email protected]> Signed-off-by: Peter Maydell <[email protected]>
# gpg: Signature made Mon 06 Feb 2017 16:13:26 GMT
# gpg: using RSA key 0xF487EF185872D723
# gpg: Good signature from "Juan Quintela <[email protected]>"
# gpg: aka "Juan Quintela <[email protected]>"
# Primary key fingerprint: 1899 FF8E DEBF 58CC EE03 4B82 F487 EF18 5872 D723
* remotes/juanquintela/tags/migration/20170206:
postcopy: Recover block devices on early failure
Postcopy: Reset state to avoid cleanup assert
vmstate registration: check return values
migration: Check for ID length
vmstate_register_with_alias_id: Take an Error **
migration: create Migration Incoming State at init time
make: move top level dir to end of include search path
The dir $(BUILD_DIR)/$(@D) was added to the include
path. This would sometimes point to a non-existant
directory, if the sub-dir in question did not contain
any target-independant files (eg tcg/). To deal with
this the rules.mak attempted to create the directory.
While this was succesful, it also caused accidental
creation of files in the parent of the build dir.
e.g. when building common source files into target
specific output files.
Rather than trying to workaround this, just revert
the code that attempted to mkdir the missing include
directories. Instead just turn off the compiler warning
in question as the missing dir is expected & harmless
in general.
NB: you can clean up a build directory parent that has
been filled with empty directories by commit ba78db44f653
using this GNU find command in that parent directory:
find audio backends block chardev crypto disas fsdev hw io linux-user \
migration nbd net qapi qom replay slirp target ui util \
-type d -empty -delete
Signed-off-by: Daniel P. Berrange <[email protected]> Tested-by: Alberto Garcia <[email protected]>
[PMM: added note about how to clean up a polluted directory] Signed-off-by: Peter Maydell <[email protected]>
Laurent Vivier [Thu, 26 Jan 2017 08:04:48 +0000 (09:04 +0100)]
linux-user: fix "apt-get update" on linux-user hppa
apt-get was hanging on linux-user hppa.
strace has shown the netlink data stream was not correctly byte swapped.
It appears the fd translator function is unregistered just after it
has been registered, so the translator function is not called.
This patch removes the fd_trans_unregister() after the do_socket()
in the TARGET_NR_socket case.
This fd_trans_unregister() was added by commit e36800c linux-user: add signalfd/signalfd4 syscalls
when do_socket() was not registering any fd translator.
And as now it is, we must remove this fd_trans_unregister() to keep them.
An early postcopy failure can be recovered from as long as we know
we haven't sent the command to run the destination.
We have to undo the bdrv_inactivate_all by calling
bdrv_invalidate_cache_all
Note that I'm not using ms->block_inactive because once we've
sent the postcopy package we dont want anything else to try
and recover the block storage on the source; the destination
might have started writing to it.
On a destination host with no userfault support an incoming
postcopy would cause the state to enter ADVISE before
it realised there was no support, and because it was in ADVISE
state it would perform a cleanup at the end. Since there
was no support the cleanup function should be unreachable,
but ends up being called and asserting.
Reset the state when we realise we have no support, thus the
cleanup doesn't happen.
Check qdev's call to vmstate_register_with_alias_id; that gets
most of the common uses; there's hundreds of calls via vmstate_register
which could get fixed over time.
The qdev id of a device can be huge if it's on the end of a chain
of bridges; in reality such chains shouldn't occur but they can
be made to by chaining PCIe bridges together.
The migration format has a number of 256 character long format
limits; check we don't hit them (we already use pstrcat/cpy but
that just protects us from buffer overruns, we fairly quickly
hit an assert).
projects / qemu.git / log