In bc1e924cf0 we've introduced video driver name and whilst
doing so we've utilized VIR_ENUM_IMPL() macro. Then, in domain
XML parsing code the generated
virDomainVideoBackendTypeFromString() is called and its return
value is assigned directly to an unsigned int variable which is
wrong. Also, the video driver enum has 'default' value which is
not formatted into domain XML but is accepted during parsing.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
The check was copied from the snapshot code and makes even less sense
here.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Semantically VIR_DOMAIN_START_AUTODESTROY doesn't really clash with
snapshot operations as the VM stays on the same host and thus bound to
the same connection. Saving the state also doesn't differ from modifying
the state of the VM which is allowed.
Remove the check as it doesn't make much sense.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Semantically we can't guarantee that we'll be able to destroy the VM on
the remote host, thus we can't allow remote migration. All other forms
of migration (e.g. saving to file) are okay though as they don't clash
with semantics of the flag.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Apart from migrating the VM to a remote host where we can't honour the
VIR_DOMAIN_START_AUTODESTROY flag properly, restricting APIs which just
modify the state of the VM does not make much sense.
Change the wording of the documentation for VIR_DOMAIN_START_AUTODESTROY
so that snapshots and saving to a file may be permitted as they
semantically don't clash with the flag itself. Otherwise we'd have to
forbid other APIs, such as virDomainDestroy as well.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
For each vhost-user GPUs,
- build a socket chardev, and pass the vhost-user socket to it
- build a vhost-user video device and associate it with the chardev
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Each vhost-user-gpu needs its own helper gpu process.
Start/stop them, and apply the emulator cgroup controller.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Call qemuExtVhostUserGPUPrepareDomain() to fill the domain with the
location of the vhost-user binary to start.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Learn to override the paths to the program to execute (vhost-user
helpers are executed to check for runtime capabilities).
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Similar to the qemu_tpm.c, add a unit with a few functions to
start/stop and setup the cgroup of the external vhost-user-gpu
process. See function documentation.
The vhost-user connection fd is set on qemuDomainVideoPrivate struct.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
See function documentation. Used in a following patch.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Add qemuVhostUserFetchConfigs() to discover vhost-user helpers.
qemuVhostUserFillDomainGPU() will find the first matching GPU helper
with the required capabilities and set the associated
vhost_user_binary.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
vhost-user device doesn't have a virgl option, it is passed to the
vhost-user-gpu helper process instead.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Check qemu capability, and accept 3d acceleration. 3d acceleration
support is checked when looking for a suitable vhost-user helper.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
To support virtio VGA with vhost-user, vhost-user-vga device is necessary.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Those new devices are available since QEMU 4.1.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
vhost-user-gpu helper takes --render-node option to specify on which
GPU should the renderning be done.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Accept a new driver name attribute to specify usage of helper process, ex:
<video>
<driver name='vhostuser'/>
<model type='virtio'/>
</video>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The same config files disovery & priority rules are used for
vhost-user backends.
No functional change, the only difference is that
qemuInteropFetchConfigs() takes a "name" argument and construct paths
with it (ex: "firmware").
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Python3 versions less than 3.7 have very unhelpful handling
of the C locale where they assume data is 7-bit only. This
violates POSIX which requires the C locale to be 8-bit clean.
Python3 >= 3.7 now assumes that the C locale is always UTF-8.
Set env variables to force LC_CTYPE to en_US.UTF-8 so that
we get UTF-8 handling on all python versions. Note we do
not use C.UTF-8 since not all C libraries support that.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The filename match rule was accidentally excluding the
Makefile.inc.am files from the long lines check.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Cleanup labels are also dropped where possible.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
Commit 7efe930ec3 introduced interlock of snapshots and checkpoints,
but the check is executed prior to the snapshot API ACL check. This
means that an unauthorized user can see whether a VM exists if it has a
checkpoint.
Move the checks to proper places.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Commit 50dfabbb59 forgot to add this important bit on how to check that
all the changes to the XML actually worked.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
A common operation in qemu_domain_address is comparing a
virPCIDeviceAddress and assigning domain, bus, slot and function
to a specific value. The former can be done with the existing
virPCIDeviceAddressEqual() helper, as long as we provide
a virPCIDeviceAddress to compare it to.
The later can be done by direct assignment of the now existing
virPCIDeviceAddress struct. The defined values of domain, bus,
slot and function will be assigned to info->addr.pci, the other
values are zeroed (which happens to be their default values too).
It's also worth noticing that all these assignments are being
conditioned by virDeviceInfoPCIAddressIsPresent() calls, thus it's
sensible to discard any non-zero values that might happen to exist
in @cont->info.addr, if we settled beforehand that @cont->info.addr
is not present or bogus.
Suggested-by: Erik Skultety <eskultet@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
A few 'cleanup' labels gone after using VIR_AUTOFREE() on the
@addrStr variable.
Reviewed-by: Erik Skultety <eskultet@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
When the bridge re-attach handling was moved out of the network driver
and into the hypervisor driver (commit b806a60e) as a part of the
refactor to split the network driver into a separate daemon, the check
was accidentally changed to only check for type='bridge'. The check for
type in this case needs to check for type='network' as well.
(at the time we thought that the two types could be conflated for
interface actual type, but this turned out to be too problematic to
do).
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
This fixes
commit b7ed8ce981
Author: Daniel P. Berrangé <berrange@redhat.com>
Date: Thu Jul 4 12:33:23 2019 +0100
remote: introduce virtproxyd daemon to handle IP connectivity
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The gnulib 'func' modules provides portability to compilers which lack
the '__func__' symbol. We only care about GCC and CLang compilers so do
not need this compatibility code.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
We removed use of the secure_getenv recently in
commit 2b0d597670
Author: Daniel P. Berrangé <berrange@redhat.com>
Date: Thu Aug 1 13:35:56 2019 +0100
util: get rid of virGetEnv{Allow,Block}SUID functions
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Use the function directly rather than having a wrapper.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Call to qemuMonitorJSONHumanCommand directly from
qemuMonitorArbitraryCommand.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
It was necessary for fallback functions but last one was deleted in
d828b744ac.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
We don't need to escape the commands any more since we use QMP
passthrough, which means we can delete the functions.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Historically HMP commands needed to be escaped to work properly.
The backdoor for calling HMP commands via QMP must unescape them so that
arguments aren't messed up.
Since we now only support the QMP passthrough the escape->unescape dance
is pointless.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
The remaining HMP commands don't require fd passing so we can purge
filedescriptor passing support from qemuMonitorJSONHumanCommandWitFd and
rename it.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
qemuMonitorHMPCommandWithFd is only called via qemuMonitorHMPCommand
macro, so we can remove the macro and the extra unused cruft from the
function.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
The handlers for 'add-fd' and 'remove-fd' are unused now and riddled
with legacy cruft. Purge them.
Last use was removed in f2019083de.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Check the disk SCSI address only when the disk actually is of
SCSI type.
Signed-off-by: Xu Yandong <xuyandong2@huawei.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The macro VIR_DELETE_ELEMENT assume that the items being deleted
have already been cleared, so we must explicitly free domain name
from the list of domains using the shared device to prevent a
memory leak.
Signed-off-by: Xu Yandong <xuyandong2@huawei.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
When registering new callback for an event, the event loop timer
must be created and registered. The timer has domain event state
object as an opaque argument which must be ref()-ed but only if
the timer was being created and registered successfully. We must
not ref it every time the virObjectEventStateRegisterID() runs.
Signed-off-by: Xu Yandong <xuyandong2@huawei.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
In function remoteDeserializeDomainDiskErrors, there is a typo.
Signed-off-by: Xu Yandong <xuyandong2@huawei.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Ján Tomko <jtomko@redhat.com>
The colors are not based on the semantics of the message but rather
on the message itself. This means that the default human-perceived
semantics (red = bad, green = good) don't really apply and spotting a
color does not mean anythting.
This is amplified by the sheer amount of output which configure produces
and the fact that some of the messages have negative semantics or
additional output.
In case of any problem the user will have to go through everything
anyways as spotting a red or yellow line has 0 information value.
Here are a few examples:
1) some 'no' messages are not a problem:
checking minix/config.h presence... no
2) some 'no' messages are actually positive:
checking for special C compiler options needed for large files... no
3) in some cases a 'yes' would mean that something is broken or needs
workaround
checking whether stat file-mode macros are broken... no
checking whether wint_t is too small... no
checking whether stdint.h predates C++11... no
checking whether the inttypes.h PRIxNN macros are broken... no
checking whether clang gives bogus warnings for -Wdouble-promotion... no
checking whether gettimeofday clobbers localtime buffer... no
4) due to string match based colors extra text makes messages yellow
checking for a traditional french locale... none
checking for working nanosleep... no (mishandles large arguments)
checking for library containing gethostbyname... none required
checking whether mbrtowc handles incomplete characters... (cached) guessing yes
5) in some cases the yes/no is very context dependant
checking whether pthread_rwlock_rdlock prefers a writer to a reader... no
checking whether this build is done by a static analysis tool... no
6) detected paths to binaries and libs are yellow despite being present
checking for objdump... objdump
checking for atomic ops implementation... gcc
As of the reasons above I don't think the colorization of the configure
output helps users or developers to debug the build process and
thus is not worth the extra code or output clutter.
This reverts commit c98174ce08.
ACKed-by: Michal Prívozník <mprivozn@redhat.com>
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
The colorization based on the string itself makes little to no sense as
the semantic meaning of the color (red = bad, green = good) is not
extracted from the semantics of the message:
1) If there is some additional string a 'yes' is marked yellow:
configure: driver_modules: yes (CFLAGS='' LIBS='-ldl')
2) In some cases a 'no' is actually good:
configure: hal: no
3) Few good/recommended configuration options are still yellow:
configure: QEMU: qemu:qemu
while using 'root:root' would still be yellow.
4) fields dumping config (e.g. the warning flags line) is a giant blob
of colored text which makes little sense
configure: Warning Flags: -fno-common -W -Wabsolute-value
-Waddress -Waddress-of-packed-member -Waggressive-loop-optimizations
-Wall -Wattribute-warning -Wattributes -Wbad-function-cast
-Wbool-compare -Wbool-operation -Wbuiltin-declaration-mismatch
-Wbuiltin-macro-redefined -Wcannot-profile -Wcast-align
-Wcast-align=strict -Wcast-function-type -Wchar-subscripts -Wclobbered
-Wcomment -Wcomments -Wcoverage-mismatch -Wcpp -Wdangling-else
-Wdate-time -Wdeprecated-declarations -Wdesignated-init
-Wdiscarded-array-qualifiers -Wdiscarded-qualifiers -Wdiv-by-zero
-Wdouble-promotion -Wduplicated-cond -Wduplicate-decl-speci ...
In addition if the idea is to switch to a more usable build system it
does not make sense to clutter the current one with more code.
This reverts commit 4b3ab5d213.
ACKed-by: Michal Prívozník <mprivozn@redhat.com>
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Use VIR_AUTO* for temporary locals and get rid of the 'cleanup' label.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
